Pain Outcomes Following Endoscopic Microvascular Decompression for Trigeminal Neuralgia Based on Vascular Compression Type

Abstract

Background  Arterial compression of the trigeminal nerve at the root entry zone has been the long-attributed cause of compressive trigeminal neuralgia despite numerous studies reporting distal and/or venous compression. The impact of compression type on patient outcomes has not been fully elucidated.

Objective  We categorized vascular compression (VC) based on vessel and location of compression to correlate pain outcomes based on compression type.

Methods  A retrospective video review of 217 patients undergoing endoscopic microvascular decompression for trigeminal neuralgia categorizing VC into five distinct types, proximal arterial compression (VC1), proximal venous compression (VC2), distal arterial compression (VC3), distal venous compression (VC4), and no VC (VC5). VC type was correlated with postoperative pain outcomes at 1 month ( n  = 179) and last follow-up (mean = 42.9 mo, n  = 134).

Results  At 1 month and longest follow-up, respectively, pain was rated as “much improved” or “very much improved” in 89 69% of patients with VC1, 86.6 and 62.5% of patients with VC2, 100 and 87.5% of patients with VC3, 83 and 62.5% of patients with VC4, and 100 and 100% of patients with VC5. Multivariate analysis demonstrated VC4 as a significant negative of predictor pain outcomes at 1 month, but not longest follow-up, and advanced age as a significant positive predictor.

Conclusion  The degree of clinical improvement in all types of VC was excellent, but at longest follow-up VC type was not a significant predictor out outcome. However distal venous compression was significantly associated with worse outcomes at 1 month.

Introduction

Trigeminal neuralgia (TN) is a clinical syndrome characterized by intense paroxysms of lancinating pain in the sensory distribution of the fifth cranial nerve. ^{1 2} TN is the most common facial pain syndrome with an annual incidence of 4.3 per 100,000 and can be caused by a variety of conditions including vascular compression (VC), multiple sclerosis, and mass lesions. ^{2 3 4} For patients with TN refractory to medical management, surgical treatment options include percutaneous glycerol rhizotomy, radiofrequency rhizotomy, mechanical balloon compression, peripheral nerve section/stimulation radiosurgery, and microvascular decompression (MVD). ⁵ Comparison of these interventions by groups such as Chang and Barbaro has demonstrated a lower rate of symptom recurrence in MVD, with an initial success rate ranging from 80.3 to 96% and long-term success rates of 84 and 74% at 1- and 10-year, respectively. ^{6 7 8 9 10 11 12 13 14 15 16 17 18} Due to its success, MVD has become the treatment of choice for younger, healthy patients with medically refractory TN. ^{11 19}

In 1967, Jannetta pioneered the use of the operating microscope to perform MVDs based on Dandy's earlier observation that TN was secondary to neuroVC. ^{20 21} With the aid of the microscope, the site of compression was initially described at the root entry zone (REZ) of the trigeminal nerve, adjacent to the pons. ^{21 22} The REZ is divided into three components: the proximal portion adjacent to the brainstem, the central myelin portion, and the central myelin–peripheral myelin transitional zone (Obersteiner–Redlich's line). ²³ While current clinic dogma suggests the REZ as the site of pathology in TN, the endoscope has enabled visualization of the distal trigeminal nerve, even in the case of large petrosal prominences, and it has been observed that compression at any location of the intracisternal nerve can result in TN. ^{24 25 26 27} While numerous anatomical and surgical studies have reported the incidence of both distal and venous compression as a source of pathology in TN, the impact of compression type on patient outcomes has not been fully elucidated. ^{6 24 28 29 30}

Given the observation that pathological compression is not limited to the REZ, it is valuable to visualize the entire nerve from the brainstem to its entrance into Meckel's cave. Compared with the microscope, the endoscope lends itself to minimal cerebellar retraction and overall improved visualization when looking around corners and boney prominences. ^{31 32 33 34 35} A comparative meta-analysis demonstrated good pain relief in 88% of patients undergoing endoscopic MVD (E-MVD) patients compared with 81% of those undergoing microscopic MVD with a recurrence rate of 9% compared with 14% with microscopic MVD and a decreased incidence of complications. ³⁶ The enhanced visualization can reveal hidden vascular loops that would otherwise be inadequately decompressed. ^{37 38 39 40}

In the current study, we performed video analysis of patients undergoing E-MVD for TN to categorize the type and location of VC along the trigeminal nerve and to further correlate pain outcomes using a reliable and validated outcome scale. ^{41 42 43 44}

Patients and Methods

A retrospective analysis of operative videos and patient outcomes was performed for patients undergoing E-MVD for TN at an academic medical center by a single surgeon (J.Y.K.L.). A 6-year interval review, from 2013 to 2019, yielded 224 consecutive patients. The start date of 2013 chosen because of institution of centralized server for surgical video backup: StorzTM Streamconnect. Patients with vertebrobasilar dolichoectasia ( N  = 2) and prior ipsilateral MVD ( N  = 5) were excluded, resulting in 217 patients. ( Table 1 )

Table 1. Patient baseline characteristics and operative data.

Variable	Overall
	( n  = 217)
Female (%)	74.2% (161)
Mean age (y)	59.9
TN type
1	86.6% (188)
2	10.6% (23)
Variable	2.8% (6)
TN pain distribution
V1	18.4% (40)
V2	64.5% (140)
V3	62.7% (136)
History of previous percutaneous rhizotomy and/or Gamma knife SRS	33.6% (73)
Side of procedure (left)	36.4% (79)
Arterial compression	77% (167)
SCA	61.3% (133)
AICA	7.8% (17)
Other artery	12.4% (27)
Venous compression	63.1% (137)
Neurolysis performed	19.4% (42)
Type I vascular compression (any proximal artery)	141 (65%)
Type II vascular compression (proximal vein without proximal artery)	36 (16.6%)
Type III vascular compression (distal artery without proximal artery/vein)	22 (10.1%)
Type IV (vascular compression distal vein without proximal artery/vein)	14 (6.5%)
Type V vascular compression (no compression)	4 (1.8%)
Number of patients available at 1-month follow-up	179 (82.5%)
Number of patients available for long-term follow-up	134 (61.8%)
Last available follow-up data (mean months)	42.9

Abbreviation: TN, trigeminal neuralgia; SRS, stereotactic radio surgery; SCA, superior cerebellar artery; AICA, anterior inferior cerebellar artery.

Prior to surgery, patients completed a reliable and validated multidimensional pain-outcome tool created by the senior author (J.Y.K.L.): the Penn Facial Pain Scale (PFPS), formerly known as Brief Pain Inventory-Facial. ^{37 38 39 40} An E-MVD decompression was performed, as described in prior publications. ^{12 45} Postoperatively, PFPS was reassessed at 1 month and at 1 year. We evaluated the absolute change (preoperative score − postoperative score) and percent improvement in two items of the PFPS as the primary clinical outcomes: facial “pain at its worst” and facial “pain at its average.” Each item was rated on an 11-point Likert scale (0/10 “no pain” to 10/10 “pain as bad as you can imagine”). In addition, pain was measured on a 7-point global impression of change (“very much improved,” “much improved,” “minimally improved,” “none,” “minimally worse,” “much worse,” “very much worse”). Long-term outcomes were determined by 1-year PFPS assessment, subsequent clinical encounters evaluating facial pain, and a phone call questionnaire.

Video analysis was conducted by first author. All 217 patients were categorized, to best characterize prevalence of VC types. In all cases, the senior surgeon visualized the cisternal trigeminal nerve in its entirety. VC was defined as any vessel making direct contact with the trigeminal nerve and categorized according to its location along the nerve. “Proximal compression” was defined as occurring from the pontine origin to the midpoint of the trigeminal nerve. “Distal compression” was defined as occurring from the midpoint of trigeminal nerve to its entry to Meckel's cave. VC was further categorized as arterial or venous. Based on these variables, VC was divided into five categories. Type 1: the most common and classic, any proximal arterial compression (alone or in combination with proximal venous, and/or any distal compression), Type 2: proximal venous compression without proximal arterial compression (alone or in addition to distal compression), Type 3: distal arterial without proximal compression (alone or in addition to distal venous compression), Type 4: distal venous compression without proximal compression, Type 5: no neurovascular conflict ( Fig. 1 ). These classifications were chosen based on classically published descriptions of VC. ^{20 21}

Fig. 1.

Endoscopic images demonstrating vascular compression (VC) types. A, artery; CN, cranial nerve; T, tentorium; V, vein; VC1, Type 1 VC; VC2, Type 2 VC; VC3, Type 3 VC; VC4, Type 4 VC.

Types 1 and 2 represent proximal compression that is well visualized with the microscope (without use of dental mirrors, drilling of the petrous tubercle or adjunct use of endoscope). Types 3 and 4 represent distal compression, which is better characterized by the introduction of the endoscope past the petrous tubercle with visualization out to Meckel's cave. Type 5 represents neurolysis only, as there was no offending vessel to decompress.

To predict pain outcomes, we constructed a multivariate model including the five types of compressive etiology, sex (binary), age (in deciles; continuous), diagnosis of TN type 1 (vs. TN type 2; binary), prior invasive intervention (MVD, radiosurgery, and/or percutaneous procedures (binary), and intraoperative neurolysis (binary). Finally, preoperative “pain at its worst/average” score was included to control for the dependence of the primary clinical outcome (absolute change). For variable selection, we performed a least absolute shrinkage and selection operator (LASSO) procedure with 10-fold cross-validation, interaction terms were not considered. The selected variables were used to construct a multivariate model. Finally, to evaluate the robustness of the selected variables, in our final multivariate model we added demographic and clinical variables identified from our literature review to be most pertinent for outcomes (if not already selected by LASSO): sex, age, and diagnosis of TN type 1. ^{6 46} All statistical analysis was performed with MATLAB (MathWorks, Natick, Massachusetts, United States). A p -value of less than 0.05 was considered significant.

Results

The majority of the patients were female with the superior cerebellar artery as the most common cause of VC, consistent with previous literature. ^{4 20 24 27} VC Type 1 (proximal arterial compression) was found in 65% of patients, Type 2 (proximal venous compression) in 16.6% of patients, Type 3 (distal arterial compression) in 10.1% of patients, Type 4 (distal venous compression) in 6.5% of patients, and no neurovascular conflict was identified in 1.8%. In 19.4% of cases (Type 1: 11%, Type 2: 33%, Type 3: 23%, Type 4: 36%) at the judgment of the senior surgeon an internal neurolysis was performed using a round knife as described by Burchiel ⁴⁷ ( Table 1 ). When possible, venous compression was decompressed via dissection and placement of Teflon/polytetrafluoroethylene (Chemours Company), similar to arterial compression types. If this was felt to be inadequate, the vein was coagulated and sacrificed.

All groups had significant improvement in facial pain scores at 1 month. In the postoperative patient global impression of change, pain was rated as “much improved” or “very much improved” in 89% of patients with Type 1, 86.6% with Type 2, 100% with Type 3, 83% of Type 4, and 100% of patients with Type 5. Additionally, percentage change in “pain at its worst” and “pain at its average” was calculated and compared with the previously described minimum clinically important difference (MCID) of 57 and 28%, respectively ( Table 2 ). ⁴¹ For all groups, the degree of improvement in both metrics was significantly greater than the MCID ( p  < 0.05), with the exception of group 5 ( p  > 0.05), which consisted only of three patients.

Table 2. Pain outcomes at 1 month.

	Overall	Type of decompression
Variable	( n  = 179)	Type 1 ( n  = 117)	Type 2 ( n  = 30)	Type 3 ( n  = 17)	Type 4 ( n  = 12)	Type 5 ( n  = 3)
(Preop) PFPS Score
Pain at its worst	8.3 ± 0.2	8.2 ± 0.2	8.4 ± 0.4	7.9 ± 0.5	8.8 ± 0.23	9.3 ± 0.5
Pain on average	6.3 ± 0.2	6.3 ± 0.2	6.2 ± 0.4	6.3 ± 0.5	6.5 ± 0.5	7.0 ± 0.7
Change in BPI Score from preop to postop
Pain at its worst (absolute)	6.2 ± 0.2	6.1 ± 0.3	6.0 ± 0.6	6.6 ± 0.4	4.9 ± 1.0	7.3 ± 1.9
% Change	73.9% ± 3.1% a	74.8% ± 3.7% a	72.6% ± 6.9% a	81.8% ± 7.0% a	56.2% ± 13.3% a	79.2% ± 20.8%
Pain on average (absolute)	5.1 ± 0.2	5.1 ± 0.3	4.6 ± 0.5	5.5 ± 0.4	4.0 ± 0.7	6.3 ± 1.7
% Change	78.0% ± 3.1% a	80.2% ± 3.4% a	78.4% ± 6.8% a	83.9% ± 5.3% a	63.5% ± 12.6% a	81.0% ± 19.0%
Patient global impression of change in pain (%)
“Much” and “very much” improved	89.4% (160)	89% (104)	86.6% (26)	100% (17)	83% (10)	100% (3)

Abbreviations: BPI, Brief Pain Inventory; PFPS, Penn Facial Pain Scale; postop, postoperative; preop, preoperative.

Notes: Percentage change in “pain at its worst” and “pain at its average” compared with minimum clinically important difference of 57 and 28%, respectively (one-sample t -test). Table values mean ± standard error of mean; absolute change = postop − preop; percentage change = ([postop − preop]/preop) × 100.

^ap  < 0.05.

The preoperative level of “pain at its worst” and “pain at its average” was similar between all groups and the degree of improvement was also similar (for “pain at its worst,” F _4,157  = 0.47, p  = 0.8; for “pain at its average,” F _4,154  = 0.69, p  = 0.6). Finally, postoperative complications were similar between all groups.

Given the overall positive outcomes and difference in the degree of improvement across groups being relatively small, we evaluated whether any compressive types were significant predictors of outcome using multivariate LASSO regression analysis. LASSO is a method to remove redundant predictors to produce a simpler model, thereby improving the generalizability of the model. Because pain was measured in a prospective manner with a validated outcome scale assessed before and after surgery, we were able to analyze several measures of pain outcomes. As such, multiple LASSO models were run to ensure robustness of the findings.

The first LASSO variable selection assessing “pain at its worst” yielded the following variables: preoperative PFPS, VC Type 4 (isolated distal venous compression), age, TN type 1, history of prior intervention for TN, and neurolysis. The variables describing the remaining four types of TN compressive etiologies were evaluated but not selected for inclusion. In the model constructed from the selected variables, VC Type 4 was a significant negative predictor of change in “pain at worst” ( p  = 0.03, Table 3 ). To assess the robustness of the predictive power of the selected variables, we then included patient sex, which was identified from our literature review to be predictive of outcomes. ⁴⁶ In the final model, VC Type 4 remained a negative predictor of outcomes. Note that we controlled for the dependence of our outcome measure (absolute change in “pain at its worst”) on the preoperative baseline severity of pain by including the preoperative “pain at its worst” in our model. Moreover, advanced age was found to be a positive predictor of outcomes ( p  = 0.003), consistent with prior literature. ⁴⁶

Table 3. Final multivariate LASSO model for “pain at its worst” and for “pain at its average”.

	Pain at its worst			Pain at its average
Variable	Regression coefficient	Standard error	p -Value	Regression coefficient	Standard error	p -Value
Age	4.9466	1.6519	0.0032038	2.6011	1.2254	0.035427
Diagnosis of TN type 1	1.0325	0.72159	0.15451	0.8976	0.52665	0.090395
Sex	0.17345	0.55621	0.75558	0.0045586	0.41603	0.99127
Prior invasive intervention	−0.74684	0.51484	0.14891	−0.53508	0.37921	0.16032
Performance of neurolysis	0.76168	0.5726	0.18542	0.71707	0.43688	0.10283
Proximal venous compression (Type 2)				−0.24764	0.49384	0.61678
Distal arterial compression (Type 3)				0.20743	0.63354	0.74381
Distal venous compression only (Type 4)	−2.0374	0.9478	0.033152	−1.4694	0.70189	0.037993

Abbreviations: LASSO, least absolute shrinkage and selection operator; TN, trigeminal neuralgia. Bold values indicate p -Value < 0.05.

We obtained congruent results in the model constructed for the outcome “pain at its average” ( Table 3 ). The variables selected with LASSO were the same as for “pain at its worst,” with the addition of VC Type 2 (proximal venous compression) and VC Type 3 (distal arterial compression). In the resultant multivariate model, VC Type 4 was again a negative predictor of outcomes ( p  = 0.037). VC Types 2 and 3 were not significant predictors of outcomes. Addition of sex for the final multivariate model did not change these results. Neurolysis was included by LASSO in both models but was not a significant predictor. Finally, we note that advanced age was again a significant positive predictor of outcomes ( p  = 0.03).

Long-term pain outcomes were obtained for 134 patients with an average follow-up of 42.9 months. Outcomes assessed for recurrence of pain and time point in which pain recurred, which is depicted by way of Kaplan–Meier curve ( Fig. 2 ). In the postoperative patient global impression of change, pain was rated as “much improved” or “very much improved” in 69% of patients with Type 1, 62.5% with Type 2, 87.5% with Type 3, 62.6% of Type 4, and 100% of patients with Type 5. Additionally, percentage change in “pain at its worst” and “pain at its average” was calculated and compared with the previously described MCID of 57 and 28%, respectively. ⁴¹ No statistically significant difference was seen between VC compression groups. A total of 74 of these patients had long-term PFPS outcomes, which also showed no statistically significant difference between VC groups noted on multivariate analysis ( Table 4 ).

Fig. 2.

Kaplan–Meier curve demonstrating long-term pain relief following endoscopic microvascular decompression for trigeminal neuralgia.

Table 4. Long-term pain outcomes.

	Overall	Type of decompression
	( n  = 134)	Type 1 ( n  = 84)	Type 2 ( n  = 24)	Type 3 ( n  = 16)	Type 4 ( n  = 8)	Type 5 ( n  = 2)
Patient global impression of change in pain (%)
“Much” and “very much” improved	70.1% (94)	69% (58)	62.5% (15)	87.5% (14)	62.5% (5)	100% (2)
Change in BPI Score from preop to postop	( n  = 74)	( n  = 49)	( n  = 13)	( n  = 7)	( n  = 3)	( n  = 2)
Pain at its worst (absolute)	4.8 ± 0.3	4.7 ± 0.3	4.8 ± 0.8	5.0 ± 0.8	4.0 ± 1.1	8.5 ± 0.4
% Change	58.4% ± 4.7%	58.2% ± 5.8%	51.5% ± 13.3%	68.3% ± 12.3%	46.3% ± 25.7%	90.0% ± 10.0%
Pain on average (absolute)	3.9 ± 0.2	3.9 ± 0.3	3.1 ± 0.6	4.1 ± 0.7	2.3 ± 0.6	7.5 ± 0.4
% Change	60.5% ± 4.2% a	61.6% ± 4.9% a	45.7% ± 12.6%	72.6% ± 10.2% a	51.7% ± 25.9% a	93.8% ± 6.3% a

Abbreviation: BPI, brief pain inventory.

Notes: Average follow-up of 42.9 months. Percentage change in “pain at its worst” and “pain at its average” compared with minimum clinically important difference of 57 and 28%, respectively (one-sample t -test). Table values mean ± standard error of mean; absolute change = postop − preop; percentage change = ([postop − preop]/preop) × 100.

^ap  < 0.05.

This study was approved by the Institutional Review Board. A waiver of informed consent was granted as this study was of minimal risk to patients.

Discussion

TN has long been attributed to compression of the trigeminal nerve at the REZ and MVD, popularized by Dr. Peter Jannetta, and remains a highly effective surgical procedure. The introduction of the endoscope has allowed for improved visualization around corners and the distal nerve even in the presence of boney prominences. Surgical decompression of the distal nerve provides unknown benefit.

We classified VC into five types to evaluate compression within and beyond the REZ and to elucidate potential differences between endoscopic and microscopic approaches. VC Types 1 and 2 represent proximal compression, identifiable with either endoscopic or microscopic approaches, and are seen in 79.8% of patients. In this context, previously reported microscopic findings of proximal arterial compression are represented by Type 1, and findings of proximal venous compression are represented by Type 2 and both represent compression at the REZ. Both groups had positive pain outcomes, significantly greater than the MCID, but neither were independent predictors of clinical outcome. Our results vary from prior reports from Barker et al, which suggested worse pain outcomes in proximal venous compression. ^{6 29} While this could be due in part to length of follow-up or statistical powering, we hypothesize that with a fully microscopic approach, the presence of additional distal compression cannot be fully assessed in all patients and may lead to inadequate decompression.

Types 3 and 4, seen in 18.5% of patients, represent distal compression beyond the REZ, most effectively identified endoscopically. We further differentiated Type 3 and 4 based on vessel type. Both groups had improved pain outcomes, significantly greater than the MCID. Type 3 was not a significant predictor of outcomes. In contrast, Type 4 was a significant negative predictor of pain outcomes at 1 month but not at longest follow-up.

This indicates that while there may be some early negative outcomes with distal venous compression, all groups have similar outcomes on long-term follow-up. Given that all groups had pain outcomes greater than the MCID, we suggest exploration and decompression of the entire cisternal segment of the nerve of both venous and arterial compression

In addition to the above findings, advanced age was an independent positive predictor of outcome. This is consistent with prior literature and should be discussed when counseling patients preoperatively. ⁴⁶

This study indicates that decompression of both proximal and distal compression by both arteries and veins can lead to positive outcomes, despite current dogma suggesting TN is secondary to proximal arterial compression at the REZ. The current hypothesis suggesting that pain is caused by demyelination of the central myelin of the REZ and subsequent hypersensitivity is not consistent with the distal compression observed in Types 3 and 4. Additionally, in the Type 5 subset of patients no neurovascular compression was observed, consistent with previously descriptions by Burchiel. ^{48 49} These findings suggest that pathology is more nuanced than isolated proximal arterial compression of the trigeminal nerve.

An alternative model proposed known as the “ignition hypothesis” postulates partially damaged neurons in the Gasserian ganglion become hyperexcitable, with spontaneous bursts and a lower activation threshold. ^{50 51 52} This hypothesis could be consistent with symptomatic compression, both venous and arterial, along the entire length of the trigeminal nerve. This suggestion has led to the investigation of ion channels, transient receptor potential cation channels, serotonin transport genes, KOR, the DREAM protein, and the Nrg/ErbB3/ErbB2 signaling complex as potential factors in the development of TN. ⁵³ While no clear etiology has been determined from these studies, continued investigation is essential for a more comprehensive understanding of TN.

This study was limited by sample size, particularly in types of VC with lower incidence and with patients being lost to follow-up in the long term. Given this, along with the significant worsened outcome of venous compression at 1 month and current literature suggesting worsened outcomes in venous compression, continued studies investigating outcomes-based compression type are warranted. Additionally, Barrow Neurological Institute Pain Intensity scale (degree of medication use) was not recorded during follow-up, thereby limiting comparison to other studies. Lastly, the rate of neurolysis (19.4%) is a potential cofounding variable, although it was not found to be significant on statistical analysis. This is made clear by patients with Type 5 compression plus neurolysis still having positive pain outcomes.

Conclusion

Since its popularization by Jannetta in 1967, MVD for TN been attributed to arterial compression of the REZ of the trigeminal nerve. We sought to use our repository of endoscopic MVD videos to classify and analysis VC types. Our study demonstrates a heterogeneity in both the location of neurovascular compression as well as the type of offending vessel. The degree of clinical improvement in all types of VC was excellent, greater than the MCID. Among these groups, distal venous compression was associated with worse pain outcomes at 1 month, but not at longest follow-up, suggesting that both distal and venous compression play a role in TN. While our understanding of TN continues to expand, there is a need for further studies to fully understand the mechanisms behind this complex pathology.