Sage Journals: Discover world-class research

Abstract

Background

Many medically-refractory trigeminal neuralgia patients are non-responders to surgical treatment. Few studies have explored how trigeminal nerve characteristics relate to surgical outcome, and none have investigated the relationship between subcortical brain structure and treatment outcomes.

Methods

We retrospectively studied trigeminal neuralgia patients undergoing surgical treatment with microvascular decompression. Preoperative magnetic resonance imaging was used for manual tracing of trigeminal nerves and automated segmentation of hippocampus, amygdala, and thalamus. Nerve and subcortical structure volumes were compared between responders and non-responders and assessed for ability to predict postoperative pain outcome.

Results

In all, 359 trigeminal neuralgia patients treated surgically from 2005–2018 were identified. A total of 34 patients met the inclusion criteria (32 with classic and two with idiopathic trigeminal neuralgia). Across all patients, thalamus volume was reduced ipsilateral compared to contralateral to the side of pain. Between responders and non-responders, non-responders exhibited larger contralateral trigeminal nerve volume, and larger ipsilateral and contralateral hippocampus volume. Through receiver-operator characteristic curve analyses, contralateral hippocampus volume correctly classified treatment outcome in 82% of cases (91% sensitive, 78% specific, p = 0.008), and contralateral nerve volume correctly classified 81% of cases (91% sensitive, 75% specific, p < 0.001). Binomial logistic regression analysis showed that contralateral hippocampus and contralateral nerve volumes together classified outcome with 84% accuracy (Nagelkerke R²= 65.1).

Conclusion

Preoperative hippocampal and trigeminal nerve volume, measured on standard clinical magnetic resonance images, may predict early non-response to surgical treatment for trigeminal neuralgia. Treatment resistance in medically refractory trigeminal neuralgia may depend on the structural features of both the trigeminal nerve and structures involved in limbic components of chronic pain.

Keywords

Trigeminal neuralgia MRI hippocampus trigeminal nerve volumetry outcome prediction

Introduction

Trigeminal neuralgia (TN) is a neuropathic facial pain disorder characterized by intermittent, typically unilateral attacks of lancinating pain in the distribution of one or more branches of the trigeminal nerve (cranial nerve V—CNV) (1). While TN is occasionally caused by lesions (e.g. demyelinating plaque in multiple sclerosis, or tumor), most cases are non-lesional and fall into two categories: a) classic TN, associated with neurovascular compression of the root entry zone (REZ) of CNV; and b) idiopathic TN, occurring in the absence of neurovascular compression (1). A substantial proportion of TN patients become refractory to medical treatment over time, and are treated with neurosurgical procedures targeting CNV directly, including microvascular decompression (MVD), percutaneous rhizotomy, or stereotactic radiosurgery. Unfortunately, pain recurrence following technically successful surgery is common: Even with MVD – the most efficacious surgical treatment for TN – early recurrence within 2 years of treatment occurs in approximately 25% of patients, with a 4% per year failure rate thereafter (2).

The presence and degree of neurovascular compression on preoperative imaging has been shown to be a predictor of positive outcome following MVD for TN (3). However, neurovascular compression of CNV in non-lesional TN is absent in as many as 28.8% of patients (4). Furthermore, a non-trivial minority of classic TN patients with prominent neurovascular compression do not achieve long-term pain relief following technically successful MVD surgery (2). To date, long-term responders to surgical treatment have not been found to have substantially different demographic features at the time of operation compared to patients who experience early pain relief (3,5). Predicting pain relief in patients undergoing TN surgery – ideally using information collected as part of a standard preoperative workup – therefore remains an unmet challenge. Solving this challenge could, in turn, improve our ability to counsel TN patients regarding prognosis following surgery, improve our ability to match patients with the best surgical treatment strategy, and ultimately to identify factors that may improve the durability of pain relief after surgery.

Neuroimaging using MRI provides a non-invasive means of generating objective biomarkers of eventual response to surgery for TN, although to date no neuroimaging metrics have consistently been able to predict postoperative pain relief. While CNV volume and cross-sectional area (CSA) – typically measured on T2-weighted MRI scans – appear to be consistently reduced on the affected side in patients with TN (6,7), the exact relationship between these simple nerve-based measures and surgical outcome has been conflicting (5,6,8 –10). Some groups have also examined microstructural alterations in CNV and its REZ with diffusion tensor imaging (DTI), and identified possible links between preoperative diffusivity metrics and pain outcomes (11 –13). However, quantitative DTI approaches are disadvantaged because their implementation requires a non-trivial degree of image processing expertise, and further because DTI acquisitions are not typically part of the routine workup in TN patients. Of note, all previous attempts at predicting long-term pain response from preoperative imaging alone have failed to take into account the potential importance of several grey matter brain regions involved in TN and other chronic pain conditions (e.g. the insula, cingulate, amygdala, and parahippocampal region, among others) (14 –17). Despite having identified that brain changes occur with successful surgical treatment of TN and other chronic pain conditions (13,18), no studies to date have looked at how preoperative brain and CNV features may interact to influence surgical outcome.

Our central hypothesis is that TN patients who do not respond to surgical treatment can be characterized by distinct neuroanatomical features – manifest both in CNV and at the brain-wide level – which distinguish them from patients who experience long-term pain relief. In the present study, we aim to identify these neuroanatomical features by performing quantitative analysis of routinely acquired clinical MRI scans obtained in TN patients prior to surgery. Specifically, we focused on differences between surgical responders and non-responders in CNV volume, as well as the volumes of three subcortical brain structures involved in trigeminal sensory relay (thalamus) or as potential contributors to limbic components of chronic pain (hippocampus, amygdala).

Methods

Study participants and data-acquisition: This was a single-centre, retrospective study of patients treated surgically for TN at the University of Alberta Hospital between 2005 and 2018, approved by the Research Ethics Board of the University of Alberta. All patients had medically-refractory classic or idiopathic TN as defined by International Classification of Headache Disorders 3rd edition (ICHD-3) criteria (1). Potential study participants were identified from an operative database, and were included if they had undergone MVD surgery performed by any one of three experienced neurosurgeons, had proof of successful technical completion of surgery according to operative notes, and underwent preoperative brain MRI scanning no more than 12 months prior to surgery. Patients with a history of multiple sclerosis or other lesional cause of TN (e.g. brain tumor), or who had undergone any previous non-TN neurosurgical procedures, were excluded. Demographic and clinical data were obtained from physical and electronic patient charts. Patients were only included in the study once, even if they had undergone multiple TN procedures. In such cases, brain imaging collected prior to the first surgical procedure, if available, was used for analysis, and response following this same initial procedure was used to define treatment outcome. If preoperative imaging was not available for the first attempted surgical treatment, preoperative imaging collected from the earliest possible surgical treatment thereafter was used for analysis, and response from this same procedure was used to define treatment outcome.

Clinical characteristics and outcome assessment: The following demographic and clinical data were collected: Sex, age at preoperative MRI, duration of TN, affected side, presence of neurovascular compression, first (virgin) surgical treatment, time to surgery, intracranial volume (ICV) and medications at time of preoperative MRI (divided by medication class, i.e. antiepileptic, antidepressant, opioid, baclofen). Duration of TN was defined as the amount of time between the date of initial TN diagnosis to the date of preoperative brain MRI acquisition, and time to surgery was defined as the amount of time between preoperative imaging and the date surgery was performed. Patients were classified as responders or non-responders according to the following criteria: Responders had documented evidence of initial pain relief following surgery, no evidence of documented pain recurrence within one year of surgery, and no evidence of repeat (or offer of repeat) surgical treatment within one year of surgery; non-responders had documented evidence of inadequate initial pain relief following surgery, or had been offered or received repeat surgical treatment within one year of initial surgery, or had documented evidence of pain recurrence following initial pain relief during that 1-year period.

Quantitative MRI analysis

Subcortical volumetric analysis: Preoperative 1.5T T1-weighted MPRAGE or SPGR MRI scans (voxel size 1 × 1 × 1 mm) without contrast were used for subcortical structural analysis. DICOM images for each patient were obtained from institutional PACS (Picture Archiving and Communication System), converted to NIFTII format, and reoriented as follows: Images from patients with left-sided TN were flipped in the axial plane with FMRIB’s FSL toolbox (19), while images from patients with right-sided TN remained in native orientation. FMRIB’s FSL (19) brain tissue segmentation tool SIENAX (20) was used to generate brain tissue (grey matter, white matter, cerebrospinal fluid) volumes and an estimate of ICV called the v-scaling factor. Bilateral hippocampus, amygdala, and thalamus volumes were determined using FSL-FIRST (21). Quality control was performed for each patient by two expert raters (authors HD and CE) who inspected all subcortical segmentations; evidence of mis-segmentation in any structure resulted in subject exclusion from all future analyses. Subcortical structure volumes were calculated for comparison in the following ways: Ipsilateral and contralateral to the side of pain, and total structure volume (ipsilateral + contralateral).

Trigeminal nerve volume analysis: Preoperative 1.5T T2-weighted CISS or FIESTA images (0.67 × 0.67 ×1 mm) were used for CNV volume analysis. Manual segmentation of CNV from its emergence at the pons to its entry at Meckel’s cave (i.e. the entire cisternal segment) was performed with the ITK-SNAP toolbox (22). CNV volume was computed from slice-by-slice tracing of the nerve in the axial plane by author EL using the Polygon Mode tool. Intra-rater reliability testing was assessed by resegmentation of CNV by the same rater in 10 random subjects 3 months following initial segmentation. Inter-rater reliability was assessed by resegmentation of CNV in 10 random subjects by a second trained rater (author SW). Intraclass correlation coefficients (ICC) were calculated using a two-way, mixed effects model, and measures are reported for average measure using absolute agreement. Nerve volumes were calculated for comparison in the following ways: Ipsilateral and contralateral to the side of pain, total nerve volume (ipsilateral + contralateral), and % difference ((ipsilateral – contralateral/ipsilateral) × 100).

Statistical analysis

Within-patient inter-side comparisons: Within-patient comparisons of ipsilateral versus contralateral volumes were performed separately for nerve and subcortical structures using repeated-measures one-way analysis of covariance (ANCOVA) [IBM SPSS Statistics for Mac OS X, version 24, IBM Corporation, Armonk, NY, USA]. Demographic variables that differed between treatment outcome groups were included as covariates (i.e. ICV and number of previous surgical procedures for TN).

Comparisons between treatment outcome groups: CNV volumes were compared between responder and non-responder groups using one-way ANCOVA. Subcortical structure volumes were compared between responders and non-responders using multivariate-ANCOVA [MANCOVA, IBM SPSS Statistics for Mac OS X, version 24, IBM Corporation, Armonk, NY, USA]. The independent variables of interest were hippocampus, amygdala, and thalamus volumes. The dependent outcome variable of interest was response versus non-response. Three separate MANCOVAs were performed, one for each of the following “sides”: Ipsilateral, contralateral, and total. Demographic and clinical variables that differed between outcome groups were included as covariates (i.e. ICV and number of previous surgical procedures for TN).

Significance levels and correction for multiple comparisons: For repeated-measures ANCOVA comparing ipsilateral versus contralateral nerve volumes, statistical significance was set at a threshold of p < 0.05. For repeated-measures ANCOVA comparing ipsilateral versus contralateral subcortical structure volumes (hippocampus, amygdala, thalamus), statistical significance was set at a threshold of p < 0.017 (i.e. 0.05/3, Bonferroni Correction for three comparisons). Bonferroni family-wise error correction was applied for between-treatment outcome comparisons of subcortical structure volumes (using MANCOVA) to generate adjusted p-values. The threshold for statistical significance for this comparison was therefore p < 0.05.

Outcome prediction: The ability of preoperative CNV and subcortical structure volumes to classify response versus non-response was assessed using receiver-operator characteristic curve (ROC) analysis (GraphPad Prism version 7 for Mac OS X, GraphPad Software, La Jolla, California, USA); the combined ability of CNV and subcortical structure volume to predict outcome was assessed using binomial logistic regression analysis (IBM SPSS Statistics for Mac OS X, version 24, IBM Corporation, Armonk, NY, USA).

Results

Study participants: We identified 359 TN patients treated with surgery between 2005 and 2018 as potential study subjects (Figure 1). Forty-two had suitable preoperative T1-weighted imaging without contrast. Five patients had a potential lesional cause of TN identified and were excluded (four multiple sclerosis and one tumor). Thirty-seven classic or idiopathic TN patients underwent FIRST automated segmentation. Quality assurance identified gross mis-segmentations in three subjects, who were excluded from further analyses. In total, 34 TN patients proceeded to subcortical volume analysis. High-resolution T2-weighted imaging is typically the first MRI acquisition performed in TN patients referred for neurosurgical assessment; as a consequence, three of the 34 patients included in the subcortical analysis had outdated T2-weighted imaging acquired more than 12 months prior to surgical intervention. Thus, only a 31-patient subgroup could also be included for CNV analysis (Figure 1).

Figure 1.

Patient selection for subcortical and CNV volumetric analysis.

Clinical characteristics and demographics: Thirty-four TN patients were included in all subcortical volumetric analyses (Table 1). Twenty-three were responders and 11 were non-responders. Thirty-two patients had neurovascular compression identified, and therefore had classic TN, while neurovascular compression was not present in the two remaining idiopathic TN patients. Responder and non-responder groups did not differ in sex distribution, mean age at surgery, duration of TN, side of pain, neurovascular compression, time to surgery, and medication classes used at the time of preoperative MRI. ICV was found to be larger on average in non-responders (p = 0.025). Additionally, a greater proportion of non-responders had previously undergone surgical treatments for TN (n = 4 instances of previous percutaneous rhizotomy in the non-responder group vs. n = 0 prior surgical treatments in the responder group; p = 0.007). Accordingly, ICV and number of previous surgical treatments were added as cofactors in subsequent analyses (Table 1).

Table 1.

Demographic and clinical characteristics of TN patients included in subcortical volumetric analysis. Values represent mean ± standard deviation where appropriate. Student’s t-test (with Welch’s correction) used where appropriate. Fisher’s exact test used for all categorical comparisons with the exception of sex (Pearson Chi-squared).

	Responders	Non-responders	p-value (two-tailed)
Outcome group	23	11	N/A
Sex (female/male)	12/11	6/5	>0.99
Age, years	52.59 ± 9.50	51.86 ± 16.7	0.90
Duration of TN, years	6.61 ± 5.07	5.55 ± 4.63	0.55
Affected side (left/right)	7/16	4/7	>0.99
Neurovascular compression (yes/no)	22/1	10/1	>0.99
Virgin surgical treatment (yes/no)	23/0	7/4	0.007*
Time to surgery, days	53.78 ± 72.0	42.09 ± 131	0.79
Intracranial volume (V-scaling factor)	1.39 ± 0.13	1.30 ± 0.09	0.025*
Antiepileptic (yes/no)	23/0	11/0	>0.99
Amitriptyline (yes/no)	1/22	2/9	0.24
Opioids (yes/no)	4/19	3/8	0.66
Baclofen (yes/no)	7/16	2/9	0.68

Denotes statistical significance (p-value < 0.05).

Time to surgery: Amount of time between pre-operative MRI and surgical treatment.

V-scaling factor: Coefficient of enlargement used to scale individual whole brains to Montreal Neurological Institute standard template.

Antiepileptic: Antiepileptic medication class: Gabapentin, pregabalin, carbamazepine, oxcarbazepine.

Opioids: Opiate medication class: Hydromorphone, oxycodone, morphine.

CNV volume: Intra-rater reliability for CNV volume was very good, with an average measures ICC 0.89 (95% confidence interval 0.54–0.97; F(9,9) = 8.15, p = 0.002). Inter-rater reliability for CNV volume was excellent, with average measures ICC 0.97 (95% confidence interval 0.86–0.99; F(9,9) = 26.15, p < 0.001). Across all patients, there was no difference in CNV volume ipsilateral and contralateral to the side of pain (37.14 ± 18.7 mm³ and 39.11 ± 18.1 mm³ respectively, p = 0.46), nor did ipsilateral and contralateral CNV volumes differ within responders or non-responders analyzed separately. Non-responders had significantly larger CNV volume contralateral to the side of pain compared to responders (53.3 ± 19.5 mm³ and 31.3 ± 11.5 mm³ respectively, p = 0.009), while there was no difference in ipsilateral and total CNV volume between outcome groups (46.2 ± 24.3 mm³ and 32.1 ± 12.9 mm³ respectively, p = 0.83; 99.6 ± 41.8 mm³ and 63.4 ± 19.6 mm³ respectively, p = 0.11) (Figure 2). There was no difference between responders and non-responders in % volume difference between ipsilateral and contralateral CNV.

Figure 2.

(a) Ipsilateral, (b) contralateral, and (c) total (ipsilateral + contralateral) CNV cisternal segment volume in responders and non-responders to surgical treatment for TN. Contralateral CNV volume is significantly larger in non-responders compared to responders (p = 0.009), while no difference in ipsilateral CNV volume or total CNV volume is observed. The influence of intracranial volume and number of previous surgical procedures was corrected using ANCOVA. Error bars represent standard error of the mean. * denotes statistical significance (p = 0.05).

Subcortical structure volumes: Across all TN patients, thalamus volume was larger contralateral to the side of pain than ipsilateral (7915 ± 633 mm³ and 7702 ± 568 mm³ respectively; p < 0.001), while no between-side volume differences were observed for hippocampus (3592 ± 373 mm³ and 3634 ± 367 mm³ respectively; p = 0.47) or amygdala (1411 ± 239 mm³ and 1352 ± 229 mm³ respectively; p = 0.08) (Figure 3). Contralateral hippocampus volume was larger in non-responders than responders (3830 ± 206 mm³ and 3479 ± 385 mm³ respectively; p = 0.032), as was ipsilateral hippocampus volume (3821 ± 274 mm³ and 3545 ±378 mm³ respectively; p = 0.012) and total hippocampus volume (7651 ± 388 mm³ and 7024 ± 668 mm³ respectively; p = 0.008) (Figure 4). There were no volume differences identified between non-responders and responders for ipsilateral amygdala (1372 ± 240 mm³ and 1342 ± 228 mm³ respectively; p = 0.91), contralateral amygdala (1499 ± 212 mm³ and 1369 ± 244 mm³ respectively; p = 0.43), and total amygdala (2872 ±410 mm³ and 2712 ± 435 mm³ respectively; p = 0.68), as well as ipsilateral thalamus (7863 ± 536 mm³ and 7625 ± 578 mm³ respectively; p = 0.49), contralateral thalamus (8140 ± 549 mm³ and 7807 ± 654 mm³ respectively; p = 0.38), and total thalamus (16003 ± 1074 mm³ and 15432 ± 1213 mm³ respectively; p = 0.42).

Figure 3.

Volumes of subcortical structures of interest for entire TN patient cohort. (a) Hippocampus, (b) amygdala, (c) thalamus. Thalamus volume is larger contralateral to the side of pain than ipsilateral (p < 0.001), while there are no between-side volume differences observed for hippocampus or amygdala. The influence of intracranial volume and number of previous surgical procedures was corrected using repeated-measures ANCOVA. Error bars represent standard error of the mean. * denotes statistical significance (p = 0.017).

Figure 4.

Hippocampal volumes in responders and non-responders to surgical treatment for TN. Hippocampal volume is larger in non-responders compared to responders (a) ipsilateral (p = 0.012) and (b) contralateral to the side of pain (p = 0.032). (c) Total hippocampal volume is also larger in non-responders than responders (p = 0.008). The influence of intracranial volume and number of previous surgical procedures was corrected using MANCOVA. Error bars represent standard error of the mean. * denotes statistical significance (p = 0.05).

Predicting surgical outcome from contralateral CNV and hippocampus volumes

Receiver-operator characteristic curve analysis: We performed ROC curve analysis to determine the ability of contralateral CNV volume and hippocampus volume to segregate surgical outcome groups (Figure 5), since statistically significant volumetric differences exist between responders and non-responders for these structures. The ROC curve generated for contralateral CNV volume and surgical outcome has an area under the curve of 0.868 (p < 0.001). The optimal operating CNV volume threshold for this model is 33.37 mm³, with 91% sensitivity and 75% specificity, correctly classifying outcome in 81% of cases (Figure 5(a) and (b)). The ROC curve generated for contralateral hippocampus volume and surgical outcome has an area under the curve of 0.787 (p = 0.008). The optimal operating contralateral hippocampus threshold volume is 3709 mm³, with 91% sensitivity and 78% specificity, correctly classifying outcome in 82% of cases (Figure 5(c) and (d)). ROC curves were also generated for both ipsilateral and total hippocampus volumes, though, these structures were not as predictive of outcome (Supplemental Figure 1).

Figure 5.

Receiver-operator curve (ROC) analysis of surgical treatment outcome in relation to contralateral CNV volume or contralateral hippocampus volume. (a) ROC curve for contralateral CNV volume and surgical outcome has area under the curve (AUC) of 0.868 (p < 0.001). The optimal operating threshold CNV volume of 33.37 mm³ has 91% sensitivity and 75% specificity for response. (b) Contralateral CNV volumes for each individual responder and non-responder to surgical treatment are displayed with optimal operating threshold volume overlaid. (c) ROC curve generated for contralateral hippocampal volume and surgical outcome has AUC of 0.787 (p = 0.008). The optimal operating threshold for hippocampus volume of 3709 mm³ has 91% sensitivity and 78% specificity for response. (d) Contralateral hippocampal volumes for each individual responder and non-responder are displayed with optimal operating threshold volume overlaid.

Binomial logistic regression analysis: A binomial logistic regression model generated for surgical outcome using contralateral CNV and contralateral hippocampus volume as predictors was statistically significant (χ²(2) = 19.9, p < 0.001), and able to reliably classify patients as responders and non-responders (Table 2). The model explained 65.1% (Nagelkerke R²) of variance in clinical outcome, and correctly classified 83.9% of cases (Table 2). Within the model, both contralateral hippocampus and contralateral CNV volumes made significant contributions to classification (p = 0.044 and p = 0.009 respectively).

Table 2.

Binomial logistic regression analysis of surgical outcome using preoperative contralateral hippocampus and contralateral CNV volume as predictor variables. The logistic regression model for hippocampus and CNV volume was statistically significant (χ²(2) = 19.9, p < 0.001), indicating that contralateral hippocampus and contralateral CNV volumes reliably classified patients as responders and non-responders. The model explained 65.1% (Nagelkerke R²) of variance in clinical outcome, and correctly classified 83.9 % of cases. Within the model, both contralateral hippocampus and CNV volumes made significant contributions to classification (p = 0.044 and p = 0.009 respectively).

Group	Correct prediction	Incorrect prediction	Percentage correct
Non-responders	9	2	81.8
Responders	17	3	85.0
Total	26	5	83.9

Discussion

We performed a single-centre, retrospective assessment of 34 TN patients (32 classic TN and two idiopathic TN) undergoing MVD surgery for TN and found that non-responders have larger average hippocampus and average contralateral CNV volume than responders. Furthermore, contralateral hippocampus and contralateral CNV volume were both good individual predictors of surgical outcome, correctly classifying 82% and 81% of cases respectively. We found that predictive capacity improved to 84% when these predictors were considered together in the same binomial logistic regression model, suggesting that both nerve and brain features may contribute to resistance to surgical treatment in TN. To our knowledge, ours is the first study where preoperative brain and CNV structure, obtained from standard preoperative clinical MRI scans, have been shown to relate to the outcome of surgical treatment for TN.

Our patients demonstrated a 68% surgical response rate, which is in agreement with previously reported literature, notwithstanding differences in how surgical outcome is measured across various studies (2). Previous work has found that CNV volume ipsilateral to the painful side of the face is typically reduced compared to the unaffected nerve in patients with TN as well as compared to healthy controls (6,7). This has been attributed to volume loss secondary to neurovascular compression of CNV ipsilateral to the side of pain (6,8). We did not find differences between ipsilateral and contralateral CNV volume across all patients. It is possible that our comparatively small sample size limited power to detect small inter-side nerve volume differences, or that neurovascular compression in our cohort was relatively non-severe. However, we did observe that contralateral thalamus volume was larger within TN patients, suggesting upstream neurons residing within the ventroposteromedial thalamus are affected in TN and likely reflect alterations in the trigeminal sensory system.

Very few studies to date have investigated the relationship between CNV volume and surgical outcomes, which was a primary focus in the present study. Previously published studies also appear to disagree on the exact nature of nerve-outcome relationships: While Cheng et al. (5) and Leal et al. (6) found that worse MVD outcome was associated with larger CNV volume and CSA ipsilateral to the side of pain, Duan et al. (8) found the opposite CSA relationship. We found no association between ipsilateral CNV volume and surgical outcomes. Rather, we identified larger average contralateral CNV volume in non-responders to surgical treatment and identified a threshold contralateral CNV volume that correctly predicted surgical outcome in 81% of cases within our cohort. Cisternal CNV volume reflects size differences across the entire cisternal nerve extent (unlike CSA); it is, therefore, well suited for capturing widespread nerve changes. Additionally, the measurement of cisternal CNV volume is less dependent on user input compared to CSA, which is measured on only a single MRI slice that needs to be selected by the observer. Our findings suggest that while initial CNV injury – say, due to neurovascular compression – may well initiate a sequence of events ultimately leading to the development of TN, delayed bilateral nerve-wide changes occurring in patients with longstanding TN might confer treatment resistance, and thus be more useful in predicting surgical outcome.

While a variety of grey matter brain structures have been implicated in TN and other chronic pain conditions (e.g. the insula, cingulate, amygdala, and parahippocampal region), exactly how cortical and subcortical brain regions may be altered in treatment resistant-TN has been under-studied (14 –17). We observed preoperative hippocampus volume to be larger in non-responders than responders and found contralateral hippocampus volume specifically to be the best individual predictor of surgical outcome, correctly predicting response in 82% of cases. While the trigeminal sensory system is certainly central to the development and maintenance of both acute and chronic pain in TN, affective limbic contributions have also been shown to be important in chronic pain experience (23,24). In humans, it has recently been shown that demyelinating lesions of the contralateral hippocampus constitute one of the most common supratentorial abnormalities in patients with TN due to multiple sclerosis (25). Additionally, a recent animal study identified a strong positive relationship between increased adult hippocampal neurogenesis and the maintenance of neuropathic pain (26). These findings are consistent with our observation that surgically-resistant TN patients have enlarged hippocampi, potentially owing to increased hippocampal activity related to the recollection and then integration of emotionally-significant stimuli with the experience of chronic pain (23,24). Specifically, why we found that hippocampus volume contralateral to the side of pain seems to be particularly important in treatment resistance remains an open question, and warrants further investigation. The phenomenon of treatment resistance in chronic pain is unlikely to be driven by a single structure, despite our findings. Future network and connectivity examinations between responders and non-responders would complement this work nicely, as the hippocampus – and potentially other limbic structures as well – may represent a node within networks working together to influence pain.

There is no doubt that abnormalities within CNV, in particular demyelination associated with vascular compression, play a key role in the development and maintenance of TN in a substantial proportion of cases (27). We speculate that our novel finding of a relationship between hippocampal volume and treatment outcome in TN may suggest that, over time, hippocampal changes also occur that may relate to hyperactivity of the trigeminal system, and may in turn explain why nerve-centered surgical approaches are not definitive (2). In this vein, Wang et al.’s recent demonstration of a correlation between CNV and brain grey matter volume supports the possibility of such a relationship (14). Indeed, the improved ability to predict surgical outcome when hippocampal and CNV metrics are considered together further underscores that both nerve and brain mechanisms may be important in TN, and in particular for the maintenance of the painful state.

Limitations

This study is not without limitations. One must be cautious in making firm conclusions from observational retrospective study designs, and future longitudinal assessments are needed to generalize our findings beyond the particular patient cohort we have studied. Another limitation of this study is that patients were classified as responders and non-responders through chart review only, raising the possibility that we may have missed non-responders lost to patient follow-up, or at worst may have misclassified them as responders. Additionally, the non-responder group may contain patients who never achieved any pain relief from surgery, as well as those who did achieve initial pain relief but then experienced early pain-recurrence; it is possible that these distinct types of patients may exhibit different structural features of the brain or CNV. Therefore, it would be advantageous that future studies utilize quantitative measures of treatment response, which would also permit accurate descriptions of response and recurrence timelines. It is also important to note that preoperative MRIs were often collected well in advance of surgery and, therefore, the brain-state we evaluated may differ from that at the time of surgery. Inter-scanner variability is another potential critique, as preoperative MRI scans were not necessarily collected on the same scanners. However, it has been demonstrated previously that FSL-FIRST (19,21) generates reliable subcortical structure volumes despite inter-scanner variation (28). Patients with right or left TN were both included in this study. As a result, brain-flipping was required in order to perform structural analysis in relation to the side of pain. We acknowledge brain-flipping as another potential limitation of our study considering hemispheric lateralization of pain processing has been demonstrated (29). While our approach may be susceptible to this particular confound, the proportion of left and right TN is equal between responder and non-responder groups, greatly increasing confidence that hemispheric lateralization is not driving our findings. Finally, the small sample size is an obvious limitation, though we did use a hypothesis-driven approach, specifically evaluating CNV and three relevant subcortical structures selected a priori, with appropriate statistical thresholds and correction for multiple comparisons.

Conclusion

We show that preoperative hippocampal and CNV volume, measured on standard clinical MRI scans, may predict early non-response to surgical treatment for TN. These findings suggest that pain-state maintenance and treatment resistance in medically refractory TN may depend on the structural features of both CNV and structures involved in limbic contributions to chronic pain.

Article highlights

Volumetric analyses applied to standard clinically acquired preoperative MR images may be useful in predicting outcome of surgical treatment for trigeminal neuralgia.

Preoperative contralateral hippocampus and contralateral trigeminal nerve volumes are larger in non-responders to surgical treatment in trigeminal neuralgia, and both are good predictors of surgical outcome.

The combination of hippocampus and trigeminal nerve volumes is the best predictor of surgical outcome, suggesting that surgical response in trigeminal neuralgia may be influenced by both trigeminal nerve and brain structure within the limbic system.

Supplemental Material

CEP877659 Supplemetal Material - Supplemental material for Hippocampal and trigeminal nerve volume predict outcome of surgical treatment for trigeminal neuralgia

Supplemental material, CEP877659 Supplemetal Material for Hippocampal and trigeminal nerve volume predict outcome of surgical treatment for trigeminal neuralgia by Hayden Danyluk, Esther Kyungsu Lee, Scott Wong, Samiha Sajida, Robert Broad, Matt Wheatley, Cameron Elliott and Tejas Sankar in Cephalalgia

Footnotes

Acknowledgements

We thank Dr. Linglong Kong for his statistical advice and assistance.

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study was supported by funds from the University Hospital Foundation, and the Edmonton Civic Employees Charitable Assistance Fund. Hayden Danyluk was also supported by the Canadian Institutes of Health Research Canada Graduate Scholarship.

ORCID iDs

Hayden Danyluk

Esther Kyungsu Lee

References

Olesen J. Headache Classification Committee of the International Headache Society (IHS). The International Classification of Headache Disorders, 3rd edition. Cephalalgia 2018; 38: 1–211.

Burchiel

KIMJ

Clarke

Haglund

. Long-term efficacy of microvascular decompression in trigeminal neuralgia. J Neurosurg 1988; 69: 35–38.

Zhang

Lei

You

, et al. The long-term outcome predictors of pure microvascular decompression for primary trigeminal neuralgia. World Neurosurg 2013; 79: 756–762.

Lee

McCartney

Burbidge

, et al. Trigeminal neuralgia occurs and recurs in the absence of neurovascular compression. J Neurosurg 2014; 120: 1048–1054.

Cheng

Meng

Liu

, et al. Nerve atrophy in trigeminal neuralgia due to neurovascular compression and its association with surgical outcomes after microvascular decompression. Acta Neurochir (Wien) 2017; 159: 1699–1705.

Leal

PRL

Barbeir

Hermier

, et al. Atrophic changes in the trigeminal nerves of patients with trigeminal neuralgia due to neurovascular compression and their association with the severity of compression and clinical outcomes. J Neurosurg 2014; 120: 1484–1495.

Erbay

Bhadelia

O’Callaghan

, et al. Nerve atrophy in severe trigeminal neuralgia: noninvasive confirmation at MR imaging – initial experience. Radiology 2006; 238: 689–92.

Duan

Sweet

Munyon

, et al. Degree of distal trigeminal nerve atrophy predicts outcome after microvascular decompression for Type 1a trigeminal neuralgia. J Neurosurg 2015; 123: 1512–1518–1512–1518.

Y-S

Lee

C-C

Guo

W-Y

, et al. Trigeminal nerve atrophy predicts pain recurrence after gamma knife stereotactic radiosurgery for classical trigeminal neuralgia. Neurosurgery 2018; 84: 927–934.

10.

Urgosik

Rulseh

Keller

, et al. Trigeminal nerve asymmetry in classic trigeminal neuralgia – evaluation by magnetic resonance imaging. Ces a Slov Neurol a Neurochir 2014; 77: 582–585.

11.

Hung

PSP

Chen

Davis

, et al. Predicting pain relief: Use of pre-surgical trigeminal nerve diffusion metrics in trigeminal neuralgia. NeuroImage Clin 2017; 15: 710–718.

12.

Tohyama

Shih-Ping Hung

Zhong

, et al. Early postsurgical diffusivity metrics for prognostication of long-term pain relief after gamma knife radiosurgery for trigeminal neuralgia. J Neurosurg 2019; 131: 539–548.

13.

DeSouza

Davis

Hodaie

. Reversal of insular and microstructural nerve abnormalities following effective surgical treatment for trigeminal neuralgia. Pain 2015; 156: 1112–1123.

14.

Wang

Yang

Cao

, et al. Correlation between nerve atrophy, brain grey matter volume and pain severity in patients with primary trigeminal neuralgia. Cephalalgia 2019; 39: 515–525.

15.

Obermann

Rodriguez-Raecke

Naegel

, et al. Gray matter volume reduction reflects chronic pain in trigeminal neuralgia. Neuroimage 2013; 74: 352–358.

16.

DeSouza

Moayedi

Chen

, et al. Sensorimotor and pain modulation brain abnormalities in trigeminal neuralgia: A paroxysmal, sensory-triggered neuropathic pain. PLoS One 2013; 8: e66340–e66340.

17.

Rodriguez-Raecke

Niemeier

Ihle

, et al. Brain gray matter decrease in chronic pain is the consequence and not the cause of pain. J Neurosci 2009; 29: 13746–13750.

18.

Rodriguez-Raecke

Niemeier

Ihle

, et al. Structural brain changes in chronic pain reflect probably neither damage nor atrophy. PLoS One 2013; 8: e54475–e54475.

19.

Smith

Jenkinson

Woolrich

, et al. Advances in functional and structural MR image analysis and implementation as FSL. Neuroimage 2004; 23: 208–219.

20.

Smith

Zhang

Jenkinson

, et al. Accurate, robust and automated longitudinal and cross-sectional brain change analysis. Neuroimage 2002; 17: 479–489.

21.

Patenaude

Smith

Kennedy

, et al. A Bayesian model of shape and appearance for subcortical brain. Neuroimage 2011; 56: 907–922–907–922.

22.

Yushkevich

Piven

Hazlett

, et al. User-guided 3D active contour segmentation of anatomical structures: Significantly improved efficiency and reliability. Neuroimage 2006; 31: 1116–1128.

23.

Shimo

Ueno

Younger

, et al. Visualization of painful experiences believed to trigger the activation of affective and emotional brain regions in subjects with low back pain. PLoS One 2011; 6: e26681–e26681.

24.

Hashmi

Baliki

Huang

, et al. Shape shifting pain: Chronification of back pain shifts brain representation from nociceptive to emotional circuits. Brain 2013; 136: 2751–2768.

25.

Fröhlich

Winder

Linker

, et al. Supratentorial lesions contribute to trigeminal neuralgia in multiple sclerosis. Cephalalgia 2018; 38: 1326–1334.

26.

Apkarian

Mutso

Centeno

, et al. Role of adult hippocampal neurogenesis in persistent pain. Pain 2016; 157: 418–428.

27.

Devor

Govrin-Lippmann

Rappaport

. Mechanism of trigeminal neuralgia: An ultrastructural analysis of trigeminal root specimens obtained during microvascular decompression surgery. J Neurosurg 2002; 96: 532–543.

28.

Nugent

Luckenbaugh

Wood

, et al. Automated subcortical segmentation using FIRST: Test-retest reliability, interscanner reliability, and comparison to manual segmentation. Hum Brain Mapp 2013; 34: 2313–2329.

29.

Symonds

Gordon

Bixby

, et al. Right-lateralized pain processing in the human cortex: An fMRI study. J Neurophysiol 2006; 95: 3823–3830.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.09 MB