A total of 33 female patients with TMDs (mean age = 32; SD = 11; range = 20 to 58) were recruited from the orofacial pain clinic at the University of Alberta (UofA) and from the local community at the UofA for an ongoing study on the effects of exercise therapy on TMD symptoms.

An experienced assessor interviewed participants and performed a physical assessment to determine eligibility. Individuals were included in the study if they were women aged between 18 and 60 years old at the time of MRI and were diagnosed with pain-related TMDs with muscle pain as a chief complaint (including both myogenous and mixed TMDs). TMDs were classified according to the new Diagnostic Criteria for Temporomandibular Disorders (DC/TMD) (45). In addition, subjects with TMDs must have experienced pain in the masticatory muscle for at least 3 months that was not affected by trauma, previous infection, or an active inflammatory cause within the last month. Participants also needed a moderate to severe baseline pain score of at least 30 mm on a 100 mm visual analog scale (VAS) (46). Exclusion criteria included pregnancy, diagnosis with any severe disease (including metabolic, rheumatoid, or vascular diseases), or any other chronic pain disorders (e.g., irritable bowel syndrome and fibromyalgia), substance abuse, abnormal neurological examination, or contraindication to MRI (e.g., metallic surgical implants). Individuals were also excluded if they were currently receiving treatment for TMDs, had received exercise therapy within 6 months before study entry, or had ever received electrotherapy (e.g., transcutaneous electrical nerve stimulation [TENS] or interferential current). Eligible participants were then informed of the nature of the research before being asked to sign an informed consent per the University of Alberta's policies on research using human subjects. The UofA ethics committee has approved this project.

We pooled scans for 33 healthy female control participants (mean age = 32; SD = 10; range = 20 to 61 years) from two studies. Nine of these controls were recruited from a similar study protocol as participants with TMDs (cohort A), while the data for the remaining 24 of the controls were obtained from a healthy aging study (47) (cohort B). The inclusion criteria from these studies were as follows: Cohort A Controls—All participants spoke English as their native or primary language, had either normal or corrected-to-normal vision, no contraindications to MRI testing, and age-appropriate scores on a non-verbal intelligence testing. Participants were excluded if they had a history of hearing impairment, stroke or neurological disorders (e.g., ADHD). Cohort B Controls—Participants had no neurological, psychiatric or developmental disabilities, significant head injuries or contraindications to MRI (neurological/psychiatric conditions include conditions diagnosed by a physician or if the participant is currently taking medication for this condition). They were recruited through advertising and provided written informed consent before study participation. Both studies were approved by the University of Alberta Human Research Ethics Board. Both studies used nearly identical protocols, with the only difference in the number of slices acquired, resulting in minor differences in repetition time (TR).

At baseline, we collected age and clinical variables from subjects with TMDs. Clinical variables (48) and questionnaires were collected using our web-protected platform (REDCAP). Patient age, weight, and height were recorded at the first MRI visit. At the first clinical visit, patients reported pain intensity using the Visual Analog Scale (VAS)-measured on a 0 mm to 100 mm scale (46) taken as an average of past-week and current pain. Additionally, patients were assessed to determine the specific TMD diagnosis and which side was the most affected (bilateral, right-predominant, or left-predominant). We also asked patients to report their medication [including selective serotonin reuptake inhibitors (SSRIs)] and whether they had other complaints, including neck pain, headache, shoulder pain, back pain, and whiplash.

Additionally, we used clinical scales to measure emotional functioning and the level of disability experienced due to TMDs. Scales included the Oral Behavior Checklist (TOBCL) (49), the Neck Disability Index (NDI) score (50), Limitations of Daily Functions (LODF) (51), the Jaw Functional Limitations Scale-20 (JFS-20) score to determine jaw function (52), the Central Sensitization Inventory (CSI) total score (53), the Patient Health Questionnaire-9 (PHQ-9) for depression (54), and the Patient Health Questionnaire-15 (PHQ-15) for somatization (45), and, finally, days in pain (within the last 6 months), days of disability interference (within the previous 30 days), characteristic pain intensity, interference score, and total disability points from the graded chronic pain scale (GCPS) (55).

To measure tenderness (pressure pain threshold [PPT]) in the masseter and temporalis muscles, physiotherapists used an algometer to apply increasing pressure to the skin above the muscle until the patient felt the first sensation of pain and signaled the therapist to remove the algometer by saying “yes” (56). PPT values are defined as the minimum applied pressure in kg/cm² that causes pain. Furthermore, based on these PPT scores, the most sensitive side of each muscle was recorded as the side that had the lowest PPT measure. We also recorded patients' self-reported side of predominant jaw pain.

The tools employed in this study have been validated and shown to be reliable and responsive methods for evaluating patients with TMDs (45). They are part of the Research Diagnostic Criteria for TMD (DC/TMD), which is the gold standard for diagnosing and characterizing patients with TMDs. In the present study, we will refer to age, weight, and height as demographic variables and the remaining measures as clinical variables. Detailed descriptions of the tools used in this study are available upon request from the authors.

MRI images were acquired at Peter S. Allen MRI Research Center at the University of Alberta on a 3T Siemens Prisma (Erlangen, Germany) using a 64-channel head coil. DTI images were acquired using multi-band spin-echo EPI sequence, with FoV 220 mm × 220 mm, matrix 148 × 148 (native resolution 1.5 × 1.5 mm² interpolated to 0.75 × 0.75 mm² in-plane on the scanner), TE = 64 ms, TR = 4,910 ms / 4,700 ms (cohort A/B), 94 or 90 slices (cohort A/B) with 1.5 mm thickness, acquired axial-oblique, parallel to AC-PC, GRAPPA factor 2, multi-band factor 2, with 6 b0 volumes and 30 directions of diffusion volumes at b = 1,000 s/mm² and b = 2,000 s/mm², in 6 min 15 s/5 min 59 s (cohort A/B). Only volumes with b = 0 s/mm² and b = 1,000 s/mm² were used for analysis. Anatomical images were acquired using T1-weighted 3D MPRAGE sequence at 1 × 1 × 1 mm³ resolution (TE = 2.21 ms, TR = 1,700 ms, TI = 880 ms, flip angle = 10 deg, GRAPPA factor 2, sagittal oblique), in 3 min 37 s (cohort A) or 0.87 × 0.87 × 0.85 mm³ resolution (TE = 2.37 ms, TR = 1,800 ms, TI = 900 ms, flip angle = 8 deg, GRAPPA factor 3, sagittal oblique) in 3 min 39 s (cohort B).

We first processed DWI images using NLSAM (https://github.com/samuelstjean/nlsam/) for stabilization (57) and de-noising (58) using all available volumes. Afterwards, only five b0 and all b1000 volumes were extracted for further analysis using ExploreDTI Version 4.8.6 (59), a graphical toolbox used to process, analyze, and visualize exploratory diffusion MRI data. Images were inspected visually for quality, looking for, but not limited to: gross subject motion, signal artifacts, signal dropouts, or missing data files. All images were of good quality and retained in the study. Preprocessing steps included signal drift correction using quadratic model, Gibbs ringing correction (5 b0; Lambda = 100; iteration = 100; step size = 0.01), registration between diffusion images and structural images, masking to remove non-diffusion weighted signal (kernel = 9; 0.5 for non-DWIs; 0.8 for DWIs), non-rigid EPI correction for distortions, and corrections for individual subject motion.

After processing the images, we performed region of interest (ROI) analyses, which generated specific tracts to extract the four diffusion parameters: FA, MD, AD, and RD. ROIs for tracking were drawn based on a template following the guidelines provided in a previous study and warped into the space of each brain for tracking in participant space (60). We used a minimum FA threshold of 0.20 and a maximum turning angle of 30 degrees to initiate and continue tracking for fiber tracking. We isolated the following tracts: cingulum near the cingulate gyrus (CGC), cingulum adjacent to the hippocampus (CGH), fornix, anterior limb of the internal capsule (ALIC), posterior limb of the internal capsule (PLIC), and uncinate fasciculus (UF) (see representative tracts in Figure 1). This tractography procedure was conducted by two trained assessors using standardized protocols as described below. A reliability analysis was conducted to determine inter- and intra-rater reliability for the two assessors. We describe this analysis in detail in the statistical analysis section below.

Tractography for the CGC and CGH was done following a previously described method (60), with a slight change. Since our images often did not capture the brainstem, we placed an “and” ROI next to the brainstem only when possible while performing tractography for the CGH. Tractography for the fornix was also done based on another protocol (61). For tractography of the ALIC, we placed three ROIs bilaterally: One “and” ROI around the internal capsule region, carefully avoiding the superior fronto-occipital fasciculus, one “not” ROI at the anterior of the splenium of the corpus callosum, and one “not” ROI approximately 13 slices inferior to the base of the “and” ROI. Additionally, we placed “not” ROIs as needed to remove any unrelated fibers. For the PLIC, we placed one “and” ROI and two “not” ROIs bilaterally. We placed the first “and” ROI at the first slice in which both the PLIC and the external capsule were clearly defined when moving the axial plane inferiorly starting at the top of the corpus callosum. We then placed two “not” ROIs: one ~10 slices anterior to the front of the “and” ROI, and the other just posterior to the splenium of the corpus callosum. Again, we placed additional “not” ROIs to remove extraneous fibers. Finally, we performed tractography of the UF following a protocol previously implemented by our lab (62).

After tractography was complete, we extracted FA, MD, AD (λ1), and RD ([λ2 + λ3]/2) values (unitless for FA and in mm²/s for MD, AD, and RD) from each tract for statistical analysis.

Participant demographics and clinical characteristics are shown in Table 1. Overall, we observed no significant differences in demographics between participants with TMDs and control participants.

FA in the right CGC was lower in subjects with TMDs than controls, while we observed the opposite relationship for the left CGH. The finding of lower FA in the right CGC is consistent with previous studies showing that the cingulate gyrus is involved in pain (67). Currently, there is a lack of literature associating WM characteristics of the CGC and CGH with TMDs, which is one area we sought to address with the present study. One study (35) of individuals with musculoskeletal pain did find a difference in FA for left CGH. However, FA for the left CGH in control subjects was higher than in participants with pain (35). Additionally, when measuring RD, they found several differences in right ALIC and right PLIC, which we did not observe (35). Another study (68) also found significantly lower FA in the cingulate gyrus and UF in patients with chronic migraines. These associations were mainly on the right side, similar to our findings. While our findings must be interpreted with caution as there were no converging findings (i.e., unlike the UF that showed several diffusion metrics were different between the groups) and there are few studies on TMDs available to which we can compare our results, it does appear that WM microstructure of the cingulum likely plays a role in chronic TMDs. Ultimately, more work is needed to better comprehend the role of the CGC and CGH in chronic musculoskeletal pain before we can make any definitive claims.

Notably, we found differences in multiple diffusivity measures between groups in the UF, with lower FA and higher MD and RD for participants with TMDs compared to controls. In general, decreased FA values may reflect a decrease in the linearity of water flow through axons and have often been associated with reduced axonal integrity (69). Additionally, MD reflects the rate of water diffusion in all axon directions, and higher MD values may relate to axonal damage and inflammation (69). While FA indicates disrupted diffusion along an axon and MD can provide a general indication of the magnitude of diffusion, AD and RD add information about the direction in which diffusion is disrupted (69). AD is a measure of diffusivity parallel to fiber tracts, whereas RD represents diffusivity perpendicular to fiber tracts. While demyelination may alter RD, it does not seem to affect AD (70), which may reflect axonal damage instead (71). Therefore, our results suggest that the right UF could have microstructural changes due to inflammation and potentially demyelination in individuals with TMDs, consistent with the finding that neuroinflammation is associated with depression and chronic pain (72). Accordingly, the UF may play an essential role in chronic pain conditions.

Another important finding in our study was the predominance of microstructural changes within the right hemisphere. These findings are intriguing because, in our cohort, most subjects with TMDs reported bilateral jaw pain, and the most sensitive side of the muscles as measured by PPT did not reveal any consistent lateralization of sensitivity. Other investigators have already observed a right-sided bias in the brain's response to pain. One study (73) that specifically investigated the lateralization of pain found strong evidence for increased pain-associated activity in the right hemisphere. This study proposes that this effect could be related to attention, which is a function strongly lateralized to the right hemisphere (74). Previous research (75) has also shown a predominantly right-hemisphere response when attending to a noxious stimulus. Therefore, it is plausible that sustained attention to pain caused by chronic pain could be at least partially responsible for the microstructural alterations seen predominantly in the right hemisphere of subjects with TMDs when compared to healthy controls.

Models including only subjects with TMDs allowed us to determine the association between demographic and clinical variables and diffusion metrics in this group. PPT values showed associations with diffusion metrics across all tracts. The right medial temporalis PPT was positively associated with FA in the model of all tracts except the fornix, and PPT for the right superior anterior masseter was positively associated with MD, AD, and RD in the model fornix. These findings are conflicting because increased PPT values indicate higher pain thresholds (decreased sensitivity). In our study, PPT values were associated with increased WM integrity as measured by FA, but decreased WM integrity as measured by the other metrics. Overall, we did not find any tract consistently associated with PPT. To our knowledge, few studies have associated diffusion tensor imaging with PPT measures. However, one study (76) of subjects with chronic neck pain and whiplash-associated disorders also did not find any association between PPT and diffusion metrics (FA, MD, and RD) in the tracts examined (CGH and tapetum).

The UF was significantly different between groups in the present study, but no clinical measures were consistently associated with diffusivity in any tracts. However, a previous study (35) found an association between FA and AD in the UF and pain intensity, severity, and catastrophizing. This disparity in findings could be due to differences in the variables measured, heterogeneity of the samples (they included men and women with a range of musculoskeletal disorders), and sample size (they had a slightly larger sample). Although we found no evidence of a relationship between UF metrics and pain outcomes, our findings still suggest that there may be UF alterations associated with chronic pain states. Since we did not measure catastrophizing, we cannot determine whether this would contribute to differences in UF diffusivity; however, pain catastrophizing is known to be related to anxiety and depression (77). As seen in our sample, SSRI use (to treat anxiety and depression) was widely associated with microstructural alterations in the fornix and contributed to the model predicting AD in all tracts except the fornix, even though the association between AD and SSRI use did not reach significance. These findings hint at a potential association between diffusion metrics and anxiety and depression but are by no means conclusive.

The present study has several strengths. One strength was the use of multilevel modeling, which allowed us to account for associations between the levels of organization of the tracts within the brains of subjects and provide more accurate estimates than would be achieved had we not accounted for the nested organization of our data (64). Additionally, we chose to explore AD and RD diffusion metrics, which are often missing in studies of musculoskeletal pain but can provide meaningful information about WM changes (34). Another strength was age- and sex-matched controls, which allowed us to identify differences between the groups while minimizing differences due to age and sex.

Although this study has many strengths, limitations must also be acknowledged. One limitation in this study is the lack of clinical variables collected from control participants. Because of this limitation, we could not determine if the associations between clinical variables and diffusion are unique to patients with TMDs or would also be found in healthy controls. However, since many of these variables are specific to chronic pain, anxiety, or depression, we expect these measures to be absent or minimal in healthy controls. Our study is also limited by the lack of homogeneity within our TMD group. While we restricted our selection criteria to include only individuals with mixed and myogenous TMDs, there may still be differences between these groups, even among individuals with varying degrees of referred pain. Our sample size did not allow us to explore sensitivity analyses by TMD diagnosis groups.

Future studies can expand on this research in various ways. For example, the longitudinal effects of TMDs are of interest. Future studies would benefit from studying the impact of chronic TMDs on the brain over an extended period. In addition, it is crucial to determine whether treatment strategies could potentially target and modify brain abnormalities seen in patients with TMDs. Our results support the growing evidence that chronic pain can alter the brain's structure and have profound emotional effects. Treatment for patients with chronic pain should focus on determining effective methods for minimizing the experience of pain in the affected areas and providing emotional support when needed.