Comparison of first-time microvascular decompression with percutaneous surgery for trigeminal neuralgia: long-term outcomes and prognostic factors

This study had patient consent and institutional ethical board approval. Clinical data for all consecutive patients who underwent surgery for TN in University Hospital Southampton were prospectively collected. Between 1996 and 2012, 622 surgical procedures were performed for TN; 185 patients who had MVD and 129 who had percutaneous surgery (54 GR, 49 TC, and 26 BC) were included in this study (Fig. 1 and Table 1). Inclusion criteria were idiopathic or classical TN without MS or tumours (appropriate investigations including MRI were performed in all patients to exclude these) [9], type 1 TN, first-time surgery for TN, and more than 12-month follow-up. Ten patients who had an MVD were lost to follow-up; 1 patient with a percutaneous procedure (TC) was lost to follow-up (P = 0.055, 2-sided Fisher’s exact test). Cases with MS or repeat surgery have been reported separately [34, 35]. All patients in this work had typical (Burchiel type I [8]) TN pain, meeting the criteria of pain in the distribution of the trigeminal nerve (Vn); an intermittent, paroxysmal course; shock-like lancinating character; light-touch triggers; and initial response to TN medications. There were no exclusions from surgical treatment on purely age or comorbidity grounds. A pooled cohort analysis of all percutaneous TN procedures in our centre has been previously published [34].

MVD, GR, and TC have been employed in our centre for over 20 years, whereas BC was introduced in 2010 after the appointment of a second neurosurgical pain consultant (GV). Patient data were collected from operation notes and clinic letters. Routine clinic follow-up of these patients was conducted at 3 months and 18 months after their operation. Additional telephonic follow-up was conducted and/or questionnaires were completed to obtain long-term data. An independent, unbiased research doctor (I.N.) or specialist TN nurse (A.L.) performed these long-term interviews, minimising potential reporting biases from patients being reluctant to disappoint their treating clinician by directly reporting poorer outcomes to them. Follow-up time extended up to 20 years (minimum 1 year; mean 51.8 months).

Initial pain relief (3 months post-procedure) was defined according to the Barrow Neurological Institute (BNI) pain intensity score [41]: ‘class I, no pain and no analgesic treatment needed; class II, occasional TN pain not requiring medication; class III, some TN pain that is adequately controlled on medication; class IV, some TN pain, not adequately controlled on medication; and class V, severe TN pain, with minimal/no relief from the procedure’. Time to recurrence after a procedure was defined as time between the operation and restarting of severe TN pain not adequately controlled by medication and/or needing re-operation. New facial numbness post-procedure was classified as none/mild, moderate, and dense, with an additional category of ‘bothersome’ numbness that was counted as a complication due to its intrusive nature. Sensory disturbance was assessed by physical examination at outpatient visits and patients’ own reports at telephone interview on how they themselves assessed the numbness.

The reporting period began before the availability of cisternography using fine-cut MRI with gradient echo sequences (3D constructive interference in steady state, CISS), which became available in our centre from the year 2000; prior to this, lower resolution MRI images with MR angiography alone were used from a 0.5-T scanner. Pre-operatively all MR films were assessed for neurovascular compression (NVC) by the consultant neurosurgeon and a consultant neuroradiologist.

All operations were performed by 1 of 2 consultant neurosurgeons (G.V. and O.S.) directly or under their personal supervision. Both surgeons were fully trained and well-experienced in all procedures prior to operating on cases for this study. The decision-making strategy we used for selecting operations for individual patients is in line with published algorithms [1, 45]. MVD was offered to all patients with TN who were medically fit and relatively young, and where the MRI scan suggested NVC. All these patients were additionally counselled for a partial sensory rhizotomy (PSR), as a ‘fall-back’ position if no convincing NVC was found at operation. When NVC was observed intra-operatively, an MVD was completed; otherwise the operation would be converted to a PSR; this occurred in only 3 cases. Where the MRI did not suggest NVC, patients were offered the option of a percutaneous lesion or (if medically fit and young) exploration with MVD or probable PSR. For patients who were less medically fit, older, and/or unwilling to undergo a larger, longer operation, a percutaneous procedure was offered. Given this, the majority of patients undergoing MVD had classical TN (with NVC demonstrated on imaging), whereas those undergoing percutaneous surgery were a more mixed population of idiopathic TN and classical TN. Patients with significant V1 pain were not offered TC due to a risk of corneal anaesthesia. All patients were counselled on all procedures, with the final choice of procedure largely dependent on individual patient preference. All patients were interviewed face-to-face in an outpatient setting by the consultant surgeon to confirm diagnosis and discuss treatment options. Information leaflets were available, as was telephonic contact with a specialist nurse who was generally present during the initial consultation.

The vast majority of patients offered MVD accepted it. A few were unsuitable due to comorbidity, and most without NVC (mainly early in the series, before the quality of MR had been improved) offered an exploration similarly agreed, expecting a partial sensory rhizotomy if exploration was negative. Patients over 85 years old were excluded for MVD as none was medically suitable. All cases up to 75 years were offered open surgery when fit, as were those up to 80 in reasonable health, and only the exceptionally fit cases aged over 80 years were offered open surgery. Many of the elderly patients however preferred a shorter ‘day-case’ percutaneous procedure, declining open surgery. Records of exact numbers of patients who were offered surgery but declined were not available; however, the senior author is confident of small numbers for those with probable NVC and medically fit who declined MVD, and for those without NVC and offered exploration, who declined this option. Similarly, we do not have records for the exact number of patients with probable NVC who were not offered MVD due to age/medical comorbidities.

Although SRS was not available at our centre, patients were referred for SRS to another centre where mass lesions of Meckel’s cave/cavernous sinus could be simultaneously treated by this modality (two patients), or critical anticoagulation ruled out a percutaneous procedure requiring even only temporary reversal (one patient). In these cases, there was no need for immediate pain relief. However, due to limited facilities for SRS in the UK during this study period and geographical distance of these from our centre, SRS was not offered as a ‘surgical’ first-line in primary TN.

Percutaneous procedures were performed under general anaesthesia or sedation with fluoroscopic guidance, with minimal variation in technique between cases. For GR, a 20-gauge needle was inserted into the medial foramen ovale following Härtel’s technique, and advanced deep toward the clivus on a lateral projection. After reversal of anaesthesia, contrast was injected with the patient in the sitting position and head flexed forwards; contrast was slowly replaced with glycerol (0.36 ml, the median volume of Meckel’s cave). The needle was then withdrawn, with patient position maintained for at least 2 h. For TC, a similar needle insertion technique was used, with the needle tip placed more laterally in the foramen ovale to reach the trigeminal ganglion. Electrode tip position was confirmed by awake testing before re-sedation and heating the electrode to between 65 and 70°C for 60 s. One thermal lesion was made, except where there was combined V2 and V3 neuralgia with evoked paraesthesiae in only one of these divisions, in which case a second lesion was performed for the other division following needle adjustment and re-testing.

Following the same needle insertion technique via the middle of the foramen ovale on its posterior rim, with a soft Fogarty balloon and minor test inflation with contrast to confirm correct balloon configuration, Vn compression entailed balloon inflation with 0.35–0.50 ml of contrast to obtain a ‘pear’ shape (known to improve pain outcomes [33]), with three alternating cycles of 60s of inflation and 60s of deflation.

For MVD, the operative technique was similar to that of published techniques [5, 18, 31]. Following a small retrosigmoid craniotomy, Vn was inspected under the microscope for vascular compression anywhere along its course. Arteries causing compression were dissected from the trigeminal root and then padded with shredded Teflon felt, or the vessels were elevated in a Teflon sling sutured to the tentorium. Patency of the superior cerebellar artery past the sling was confirmed using either micro-Doppler or, more recently, intra-operative indocyanine green angiography. Compressive veins were coagulated and divided. No adverse events followed vein coagulation. The veins divided were usually transverse, draining to the superior petrosal vein, or in one or two, the superior petrosal vein itself. Neurovascular compression of Vn was noted intra-operatively using a published grading system [27]: grade I, mere contact with Vn; grade II, displacement or distortion of the nerve root; and grade III, marked indentation of Vn.

Statistical analysis was performed using IBM SPSS 25 (IBM Analytics, New York) software, with a significance level of P < .05 for all tests. The 2-sided chi-square test or Fisher’s exact test was used to compare proportions between groups of patients. Kaplan-Meier curves were compared by the log-rank test. Data were censored if there was no pain recurrence at the most recent follow-up. To identify predictors of pain relief, univariate analysis was performed by logistic regression; multivariate logistic regression was performed using those variables exhibiting a P < .2 from univariate analysis.

For MVD, 87.0% of operations yielded class I or II pain relief at 3 months, in comparison with 67.2% of patients who had a percutaneous procedure. Only 7.0% of patients who had undergone MVD and 11.7% of patients who had undergone a percutaneous procedure had little or no relief after their operation (BNI class IV and V). Comparing the distributions of initial pain relief demonstrated that MVDs provided significantly better initial relief compared with percutaneous procedures (2-sided chi-square test, P < .001; Fig. 2a).

Subgroup analysis showed that initial class I or II relief rates for GR, TC, and BC were 63.0%, 62.5%, and 84.6%, respectively. Little or no pain relief (class IV or V) was achieved after 16.7%, 10.4%, and 3.8% of GR, TC, and BC procedures, respectively. MVD provided significantly better initial relief than GR and TC (P < .001 for both MVD vs GR and MVD vs TC, chi-square test) but not compared with BC when comparing the distributions of pain relief (P = .45, chi-square test). There were no significant differences in distributions of initial pain relief scores between GR, TC, and BC (GR vs TC, P = .68; GR vs BC, P = .21; TC vs BC, P = .20, chi-square test; Fig. 2b).

Following MVD, there were 3 (1.6%) cases of CSF leak requiring temporary CSF diversion with a lumbar drain, 8 (4.3%) cases of minor stroke (typically some hemisensory loss and mild weakness that improved), 2 (1.1%) cases of chemical meningitis that recovered, and 1 (0.5%) case of deep venous thrombosis (DVT). Two (1.1%) patients reported facial paraesthesiae and 7 (3.8%) reported dysaesthesiae. There was one (0.5%) death after an MVD due to a post-operative posterior fossa haematoma in a patient with underlying but peri-operatively treated coagulopathy. The commonest complications were minor: ataxia (9.7%) and diplopia (5.9%) which improved (Fig. 3a).

After a percutaneous procedure, there were 15 (11.6%) cases of minor dysaesthesiae and 3 (2.3%) of major dysaesthesiae, 3 (2.3%) of bothersome numbness, 12 (9.3%) of reduced corneal reflex, and 9 (7.0%) of facial paraesthesiae (Fig. 3b). There were no deaths or cases of anaesthesia dolorosa. Complications stratified by type of percutaneous procedure are indicated in Supplementary Table 1. The overall complication rate (including minor and transient complications) for MVD was 24.9% and for percutaneous procedures was 35.7% (P = .04, Fisher’s exact test).

At 3-month follow-up following MVD, 4.3% of patients had mild to moderate facial numbness, and 1.6% had dense numbness; there were no cases of bothersome numbness. In contrast, 90.4% of percutaneous cases developed mild to moderate numbness, and there were significantly more cases of dense (6.1%) and bothersome (2.6%) numbness than with MVD (P = .006, Fisher’s exact test; Fig. 4a).

Subgroup analysis of percutaneous procedures showed that BC generated a greater proportion of cases with dense or bothersome post-operative numbness than GR and TC (23.1% in BC, 5.0% in GR, and 4.2% in TC; BC vs GR, P = .05; BC vs TC, P = .02, Fisher’s exact test; Fig. 4b). There was no significant difference in the proportion of cases with dense/bothersome numbness with GR compared with TC (P = 1.00, Fisher’s exact test).

Kaplan-Meier (KM) analysis for time to recurrence of TN demonstrated that MVD provided significantly longer pain relief compared with percutaneous procedures overall (P < .001, log-rank test; Fig. 5a). The median time to pain recurrence for percutaneous procedures was 36 months (95% confidence interval, CI, 24.9–47.1 months) and was not yet reached for MVDs; 25% of patients had pain recurrence at 96 months following an MVD compared with only 12 months after a percutaneous procedure.

Subgroup analysis confirmed that the durability of pain relief for MVD was significantly better compared with GR and TC (P < .001 for both MVD vs GR and MVD vs TC, log-rank test), and non-inferior to BC (P = .27, log-rank test). The median time to pain recurrence was 29 months (95% CI 16.1–41.9 months) for GR, 30 months (95% CI 8.6–51.4 months) for TC, and not yet reached for BC. Twenty-five percent of patients had pain recurrence at 5 months following a GR, 12 months after a TC, and 26 months after a BC. Although there was no difference in durability of relief after a GR compared with TC (P = .24, log-rank test), BC provided significantly longer pain relief compared with GR and TC (P = .011 for BC vs GR; P = .031 for BC vs TC, log-rank test; Fig. 5b).

For MVD, age, sex, pre-operative duration of TN pain, and trigeminal division of pain were not significant predictors of favourable outcome after surgery (as defined by initial post-operative BNI score I or II and no recurrence on long-term follow-up) using univariate analysis. With pre-operative MRI findings of neurovascular contact or compression of Vn, there was a favourable outcome in more cases (P = .10, OR 2.48, 95% CI 0.83–7.42), and similarly for severe intra-operative NVC (grade II or III; P = .06, OR 2.12, 95% CI 0.97–4.65) on univariate analysis (Table 2); these differences were not statistically significant. The type of compressive blood vessel (artery or vein) was not associated with outcome. Multivariate logistic regression analysis revealed no significant predictors of outcome after MVD. The causative vessel compressing Vn was observed to be the superior cerebellar artery in 67.2%, anterior inferior cerebellar artery in 4.5%, multiple arteries in 8.1%, an artery and a vein in 11.6%, and a vein only in 8.6% of cases.

Univariate logistic regression confirmed that, for all percutaneous procedures combined, sex, side and distribution of TN pain, and pre-operative duration of TN symptoms were not predictors of outcome. Older age and the presence of post-operative dense or bothersome numbness were the only significant predictors of favourable outcome (P = .007, OR 1.06, 95% CI 1.02–1.10 for age; P = .02, OR 5.17, 95% CI 1.27–21.12 for post-operative numbness). On multivariate logistic regression analysis, both age and post-operative numbness remained significant predictors of outcome (P = .01 and .03, respectively; Table 3).