Temporal Summation of Heat Pain in Temporomandibular Disorder Patients

Abstract

Aims

One possible mechanism underlying myofascial temporomandibular disorders (TMD) is altered central nervous system processing of painful stimuli. The current study aimed to compare TMD cases to controls on two measures of central processing, i.e., temporal summation of heat pain and decay of subsequent aftersensations, following thermal stimulation in both a trigeminal and extratrigeminal area.

Methods

Using a “wind-up” (WU) protocol, 19 female TMD patients and 17 controls were exposed to 15 heat stimuli at a rate of 0.3 Hz. Numeric pain ratings were elicited after the 1^st, 5^th, 10^th and 15^th stimulus presentation and every 15 seconds after final presentation (aftersensations), for up to 2 minutes. In separate trials, the thermode was placed on the thenar eminence of the hand and the skin overlying the masseter muscle.

Results

Groups did not differ with respect to the slope of WU when stimulated at either anatomic site, although asymptote occurred sooner for TMD patients than controls. In analysis of aftersensations, a significant group x site x time interaction was detected, in which TMD patients experienced more prolonged painful aftersensations than controls when stimulated on the skin overlying the masseter muscles.

Conclusions

These results are consistent with the presence of enhanced central sensitivity in TMD and suggest that this sensitivity may be largely confined to the region of clinical pain. This contrasts with conditions such as fibromyalgia, where central sensitivity appears to be widespread.

Temporomandibular disorders (TMD)¹ are a heterogeneous group of conditions affecting the temporomandibular joint and masticatory muscles, in which the predominant subtype^{2, 3} involves masticatory muscle pain. The cause and appropriate treatment of TMDs remain controversial.

As part of the search for underlying mechanism, research has examined whether TMD patients have altered responses to experimental pain stimuli. Such research has reached mixed conclusions. When measuring pain thresholds, some studies have shown that, compared to pain-free controls, TMD patients have lower thresholds to thermal and pressure stimuli at both masticatory sites^4–6 and remote regions.^7–11 However, some studies^{6, 9, 12–16} found no differences between TMD subjects and controls in perception of heat or pressure stimuli at extratrigeminal sites. For example, a relatively early study¹⁵ found normal thermal nociceptive thresholds in patients with masticatory muscle pain, in both the masseter and forearm areas. More recently, Svensson et al.⁵ found higher responsiveness of masticatory muscle pain patients to deep phasic stimulation in an extra-trigeminal area, but not to tonic stimulation, suggesting that altered pain processing was primarily confined to the trigeminal region. We are not the first to conclude that the evidence for altered pain processing in TMD patients is mixed.¹⁷

Wind-up

Wind-up (WU), or its psychophysical correlate, temporal summation of second pain, results from repetitive stimulation of nociceptive C-fibers.^{18, 19} In humans, WU can be measured using standard psychophysical techniques and are characterized by a progressive increase of perceived pain, hyperalgesia and enlargement of receptive fields in response to repetitive application of a constant pain stimulus.^{20, 21} WU is dependent on stimulation of N-Methly D-Aspartate (NMDA) and neurokinin (NK1) receptors that in turn sensitize wide dynamic range (WDR) and nociceptive specific (NS) neurons within the dorsal horn of the spinal cord.¹⁹ Therefore, it is a measure of central synaptic plasticity. WU can serve as both an amplification and maintenance mechanism for pain and central sensitization.¹⁹ However, WU and central sensitization are not the same phenomenon.²² Nevertheless, because WU is mediated by central (spinal) mechanisms, it can be used in human studies to determine the degree of CNS excitability to nociceptive stimuli.^{19, 22} This is important, because several lines of evidence suggest that chronic musculoskeletal pain conditions may be maintained by central nervous system dysregulation of endogenous pain control. Specifically, multiple studies (see ²³ for review) have found evidence of abnormal WU and slower dissipation of painful aftersensations in patients with fibromyalgia, a widespread pain condition that is comorbid with TMDs.^24–30

At least two groups of researchers found greater WU in response to thermal stimuli, when TMD patients were compared to normal controls,^{4 31} implying that TMD is associated with sensitization of general pain regulatory systems. Case-control differences in thermal WU patterns were found, when applying the stimulus to the glabrous surface of the hand⁴ or the fingers.³² The latter study ³² also reported more intense and frequent painful aftersensations, following termination of the thermal pulses. A focus on WU aftersensations is important to note, because this response is not dependent on the afferent stimulus and is, therefore, more easily interpretable as an indicator of central sensitivity.

In an intriguing longitudinal study, Maixner et al^33–35 presented early findings from study of a three-year cohort of initially TMD-free young women. Baseline C-fiber mediated thermal WU was associated with the development of TMD symptoms, but not at a statistically significant level.³⁵ Since cross-sectional findings sited earlier ^{4, 31} cannot sort out whether elevations in WU in TMD patients are a cause or consequence of symptoms, these longitudinal findings are arguably the most provocative findings to date supporting WU and possibly central sensitization as risk factors for the onset of TMD pain.

To date, no previous research has assessed thermal WU, applying the thermal stimulus to both the face and an extratrigeminal region such as the hand. This investigation aims to (1) Demonstrate the feasibility of eliciting a WU response in the trigeminal region (2) Compare WU responses and aftersensations generated in both an extratrigeminal and trigeminal region, (3) Compare TMD cases to controls on WU responses and aftersensations generated in trigeminal and extratrigeminal regions, and (4) provide data on the effect of modifying criteria for the definition of a WU response on both the ability to elicit a WU response and the ability to detect differences between clinical groups and anatomic sites over time.

Methods

Subject selection

TMD patients were recruited from the orofacial pain clinic at New Jersey Dental School following institutional review board approval at the University of Medicine and Dentistry of New Jersey. In order to be eligible, the treating orofacial pain specialist had to diagnose the patient with a TMD and make a clinical judgment that the pain was primarily of muscle origin. Thus, patients were most likely to fulfill criteria for a myofascial subtype of TMD.¹

Control volunteers without self-report of chronic facial pain were recruited through flyer postings on the academic campus. Although eligible controls could not have facial pain, they were permitted to have other painful disorders, since selection of pain-free controls can lead to problematic bias and error. ³⁶

Recruitment was restricted to adult women between ages 18–65, given that pain threshold and tolerance may differ for women and men. All subjects were English-speaking. Both groups of participants received a $50 honorarium.

Self-report questionnaires

Respondents completed several self-report measures. They were asked to place an ‘X’ at each location on an anatomic map of the head and neck and entire body (front and back), to indicate locations in which they had pain in the past two weeks. Time of onset of pain (months) was assessed, and participants rated their current, worst and average pain (past 6 months) on a numerical 0 to 10 scale, where “0” equaled “no pain” and “10” equaled “pain as bad as it could be.” Demographic information, including age, education and race, was collected. As an exploratory measure, the State-Trait Anxiety Inventory (STAI) was administered. The STAI is the most widely used self-report instrument for measuring anxiety in adults,³⁷ in which two 20-item self-report scales assess anxiety-proneness (trait) and current anxiety (state) level.

Sensory testing

We measured subjective responses to heat pain at the masseter muscle (bilaterally) and the thenar eminence of the non-dominant hand, using standardized heat stimuli (Medoc TSA2001 Thermal Sensory analyzer with a 900mm² Peltier thermode; Medoc Advanced Medical Systems, Minneapolis MN)

Heat pain was chosen because it can be delivered in a standardized fashion, identical stimuli can be presented to participants, and the rate of perceptual WU to heat pain has been shown to be independent of skin temperature changes that accompany repeated contacts of a heated thermode.³⁸ Heat pain thresholds were measured using a double random stair case method. The stimuli were applied to each site, starting from a baseline temperature of 32° C. Temperature within each staircase increased at 1°C/sec in 2° steps until the subject reported that the stimulus increment in that interval was painful. Once pain was indicated, the temperatures were adjusted first downward and then upward in 1° and then in 0.3° steps, until a consistent threshold could be computed.

Our primary psychophysical pain outcome was related to perceptual WU. We employed a WU paradigm similar to that successfully employed in several investigations of fibromyalgia.^{21, 39, 40} Temporal summation of pain sensations was evaluated by applying a preheated thermode to the thenar eminence of the non-dominant hand and to the skin overlying the body of the right and left masseter muscles.

The subject held the thermode assembly against either her face or hand. Contact of the preheated thermode with the skin was controlled by a computer linked to a step motor that physically moved the thermode to and from the skin. The step motor/thermode assembly was mounted in the emptied housing of a 6-volt flashlight. A plexiglass plate, cut out in the middle to allow passage of the thermode, closed off the open end of the housing. Without power, the thermode sat approximately 1 cm below the level of the plate. When energized, the thermode moved into the opening, level with outside surface of the plate. The skin tended to fall below the surface of the opening in the plate, so that the energized probe made full contact with the skin surface, if subjects simply kept the plate lightly touching the skin, as they were instructed to do for the entire 45 sec required to deliver each of the WU stimulus trains.

The probe was mechanically cycled to contact the skin for 1 sec and to dwell away from the skin for 2 sec (0.3 Hz). A total of 15 stimuli were presented.

The thermode temperature was determined from the method described by Staud et al.⁴¹ Briefly, this temperature was determined (separately for each site) through an iterative procedure as that which elicited a numerical pain-scale response of 20 or less to a first stimulus and a numerical pain-scale response that was at least 25 points greater in response to the 15^th stimulus. In this way, a subjectively equal WU response was elicited from each subject. This procedure is preferred over a constant stimulus as it minimizes withdrawal of sensitive subjects ³⁹ and provides clear interpretation of aftersensations, considered the primary indication of central sensitivity.

Subjects rated their experience using a numerical pain scale (NPS) previously validated to measure WU in FM.^{21, 39} The scale includes verbal anchors set at 10 point increments, ranging from 10 = warm (no pain) to 100=intolerable pain (20= just painful). Subjects were instructed to attend to and report intensity of the peak pain, and that peak pain would be expected to occur 1–2 seconds following stimulus presentation. They were also told that the thermal sensations may or may not be painful, and to rate each stimulus using the numerical pain scale. Subjects were cued to report on the 1^st, 5^th, 10^th and 15^th stimulus presentations, and also cued to rate the intensity of residual sensations at 15s intervals for 2 min after the 15^th stimulus (WU aftersensations).

Statistical analysis

SPSS version 14 (SPSS Inc., Chicago, IL) was utilized for statistical analysis. Group differences in pain thresholds were tested by use of the Student’s t-test. Modeling the development of WU and the decay of aftersensations employed an unstructured mixed model approach detailed by Singer and Willett ⁴². This analysis is conceptually similar to repeated measures ANOVA, but makes no assumptions about compound symmetry, and allows for unbalanced designs. This latter feature is crucial to our ability to analyze current data, in which some subjects are missing WU data at one or more sites, and some subjects have multiple WU trials at some sites. The analysis employed full maximum likelihood (ML) estimation, which produces models that maximize the likelihood of both sampling errors (like restricted maximum likelihood (RML)) and parameters (unlike RML). Inclusion of the latter makes possible the comparison between non-nested models. Significance indicates that the probability of a Type I error was no more than 5%.

Results

Table 1 summarizes the clinical and demographic characteristics of the sample. There was no significant difference between patient and control groups in either demographic characteristics or in pain thresholds on either the face or the hand. Patients did, however, report higher levels of both State and Trait Anxiety, and reported more pain outside the facial region, particularly in the axial region. These reports were focused on neck or shoulder pain (data not shown). About half of the patients reported episodic pain with an average intensity of 5 out of 10, suggesting moderate pain. The median duration of pain was 12 months.

Table 1.

Demographic, clinical and psychophysical characteristics of the sample.

Measure	Control (n= 17)	TMD (n= 19)	X2 or t	p-level
Age (yr) M (SD)	37.5 (13.8)	36.2 (13.1)	0.3	.78
Race			2.97	.23
White	6 (37.5%)	9 (52.9%)
Black	6 (37.5)	2 (11.8)
Asian	4 (25.0)	6 (35.3)
Hispanic	1 (6.3)	5 (26.3)	3.4	.18
Ever married	8 (47.1)	11 (57.9)	2.9	.41
State anxiety M (SD)	25.6 (9.1)	31.4 (5.6)	2.3	.03
Trait anxiety M (SD)	31.4 (6.1)	37.5 (6.8)	2.8	.008
Face pain (left)	1 (5.9)	18 (94.7)	28.4	<.001
Face pain (right)	1 (5.9)	17 (89.5)	25.1	<.001
Body pain
Upper extremity (left)	0	2 (10.5)	1.9	.17
Upper extremity (right)	2 (11.8)	4 (21.1)	1.7	.19
Lower extremity (left)	2 (11.8)	5 (26.3)	1.2	.27
Lower extremity (right)	3 (17.6)	4 (21.1)	.1	.80
Axial	2 (11.8)	14 (73.7)	13.9	<.001
Total painful sites (of 5) M(SD)	0.5 (0.8)	1.5 (1.5)	2.7	.01
Clinical pain characteristics
Persistent / Recurrent	--	44.4/55.6%
Pain Intensity M(SD)
Present	--	4.3 (2.2)
Worst last 6 mo	--	6.8 (2.6)
Average last 6 mo	--	5.1 (2.4)
Median pain duration (mo)	--	12
Pain threshold (°C)
Thenar eminence M (SD)	46.4 (3.6)	45.8 (2.9)	0.6	.56
Maxillary skin (left) M (SD)	43.6 (4.2)	42.5 (4.2)	0.8	.45
Maxillary skin (right) M (SD)	42.5 (4.1)	41.8 (4.1)	0.5	.59
Provides at least 1 WU trial*	12 (70.6)	10 (52.6)	1.2	.27

^*A WU trial is defined as one in which the report to the first tap was less than or equal to 20, and the report to the last tap was at least 25 points higher than the first.

A statistically similar proportion of controls and patients, ~50–70%, provided at least one valid windup trial. That is, a temperature was found which, when repeated every 3 seconds, produced a NPS response to the first stimulus of 20 or less, and a response to the fourth rated stimulus at least 25 points greater. In order to find a WU stimulus, up to 16 attempts were made on the face and up to 13 attempts were made on the hand, over the course of a testing session lasting a maximum of two hours. On average, finding or repeating a WU stimulus required 9.8 trials (SD= 3.0) on the face and 5.9 (SD=2.1) on the hand. Twenty-two of the 36 subjects provided at least one WU trial, and 14 provided 2 or more; 10 subjects provided both a face and hand trial, while another 12 provided only a hand trial. The ability to elicit a WU trial on either the hand or the face was statistically similar in patients and controls (details available upon request).

Table 2 details the effect of changing the criterion for defining a trial as one showing WU on the number of subjects that could be included in the analysis, and the total number of trials that would have met criterion. Relaxing criteria, either by reducing the necessary increase in the NPS rating or by allowing the response to the first stimulus be above the minimum report of pain, had the effect of increasing the number of subjects eligible for analysis as well as the number of trials meeting criterion. Compared to the most stringent criterion that was employed in analyses below, the most relaxed criteria would have increased the number of eligible subjects by 50% and tripled the number of eligible trials. Nevertheless, principal analyses focus on data produced from the most stringent definition, even at the expenses of reduced statistical power, as it was considered most interpretable (i.e., change in sensation from nonpainful to painful).

Table 2.

Effect of varying criteria for defining a WU trial on the number of subjects and trials eligible for analysis.

Criterion	N (%) included	Trials @ criterion
Tap1 ≤ 20 and Δ ≥ 25	22 (61.1)	80
Any Δ ≥ 25	22 (61.1)	103
Tap1 ≤ 20 and Δ ≥ 20	25 (69.4)	98
Any Δ ≥ 20	25 (69.4)	133
Tap1 ≤ 20 and Δ ≥ 15	27 (.75)	123
Any Δ ≥ 15	28 (77.8)	181
Tap1 ≤ 20 and Δ ≥ 10	30 (83.3)	147
Any Δ ≥ 10	33 (91.7)	235

Figure 1 shows the report of WU sensation during stimulation (1–45 sec) and following stimulus termination (60–165 sec, aftersensations). Inspection shows similar trajectories during stimulation in patients and controls and in the hand and face during stimulation, although patients appear to give higher ratings at 30 sec. Similar trajectories were also seen in the slope of aftersensations, but appeared most shallow when patients were stimulated on the face. Comparison of these slopes is presented in analyses below, as Figure 1 does not control for either stimulus temperature or for the presence of multiple windup trials in some subjects.

Figure 1.

Numeric Pain Scale (NPS) ratings after 1^st, 5^th, 10^th and 15^th heated thermode presentation and every 15 sec of aftersensations (no thermode presentation)

Note: Error Bars shown only for Control (C) subject hands, for purposes of visual presentation.

WU response

As seen in Fig. 1, and by definition, responses increased during successive repetitions of the WU stimulus (1–45 sec). Mixed models analysis showed an intercept of 18.4 (SEM= 1.2) and a regression coefficient of 12.1 (SEM= 0.6) attributable to TIME, indicating an average initial response of about 18 NPS units and an increase of about 12 NPS units in each successive report interval. Subsequent inclusion of temperature, anatomic site and patient group into the model failed to improve its fit, indicating that this slope was similar for the hand and face, patient and control group, and for various WU temperatures. While the overall test of the interaction between group and time was not significant (F(3, 307)= 1.81, p= .14), post-hoc tests showed higher NPS ratings in patients than controls at 30 sec (M(SD)= 46.8 (9.6) vs. 39.7 (9.3), p< .05), but not at 45 sec or at earlier time points, supporting the observation in Fig 1 that patients reach asymptotic levels sooner than controls.

Secondary analyses compared subject groups and anatomic sites on the temperature necessary to elicit a WU response. Mixed model analysis showed that, while WU on the face required a lower temperature than on the hand (M(SD)= 49.6 (2.3) vs. 47.8 (2.7), p< .001), there was no difference between patients and controls and no interaction of group by site. Thus, although there was no difference in WU temperature attributable to patient status, temperature required to elicit a WU response was lower on the face than the hands of the average subject.

Modeling of aftersensations

Model 1, Unconditional Means (UM)

Table 3 shows that the intercept (the average initial post-stimulation reported sensation) over all persons and all post-stimulation intervals was 14.9, indicating that after-sensations were not, on average, painful, but were significantly greater than zero. Inspection of the variance components suggests non-zero levels of both within and between-person variance, justifying the search for variables to which this variance may be attributed. The ICC was .71, indicating that 71% of the variability among reports of aftersensations was attributable to differences between persons. The ICC is also interpreted as the autocorrelation among residuals in the repeated observations. Because it is significantly different from zero, this mixed model provides a more appropriate analysis of these data than would be obtained with a repeated measures ANOVA, whose results are valid only when these errors are uncorrelated.

Table 3.

Results of mixed model analysis of initial status and rate of change of aftersensations.

Initial Status	Model 1	Model 2	Model 3	Model 4
Intercept	14.9 (2.75)a	21.4 (2.84) a	21.9 (2.7) a	22.6 (3.8) b
Group				−1.7 (5.4)
Site				1.1 (1.6)
G X S				−.63 (1.93)
Temp			.96 (.19) a	1.26 (.24) a
Rate of Change
Time		−1.87 (.32) a	−1.87 (.32) a	−1.81 (.49) a
Group x Time				−.16 (.68)
Site X Time				−.58 (.36)
G X S X T				1.33 (.46) b
Variance Components
Level 1 (within S)	67.1 (3.71) a	35.9 (2.05) a	34.6 (1.98) a	33.0 (1.88) a
Level 2 (between S)
Initial status	162.3 (50.62) a	170.5 (54.3) b	152.9 (43.9) b	149.0 (47.6) b
Rate of change		1.82 (.75) b	1.85 (.75) c	2.08 (.81) c
Covariance		−3.9 (4.6)	−2.36 (4.32)	−2.02 (4.44)
−2 LL	4762.9	4404.0	4380.1	4350.8
AIC*	4768.9	4416.0	4392.1	4376.8
BIC**	4782.4	4442.0	4419.1	4435.3

^ap< .001

^bp< .01

^cp< .05

^*, Aikike’s Information Criterion;

^**, Bayesian Information Criterion

Model 2, Unconditional Growth (UG)

Adding TIME to the UM model shows the first post-stimulation report to be slightly painful, i.e., 21.4. The change in report as a function of post-stimulus TIME was significantly different from 0, indicating an overall change in the level of aftersensations during the post-stimulation interval. The magnitude and sign of the regression coefficient, (-1.83) indicates that average report levels decrease by about 2 NPS units for each 15-second period of the post-stimulation interval. Inspection of the variance components shows that inclusion of TIME reduces both level 1 (within person variability in true trajectory of change) as well as level 2 variability (between persons). In the first case, a pseudo-R² of .465 indicates nearly a 50% reduction in residual variability at level 1, and a 20% reduction at level 2.

Model 3, add temperature covariate to UG model

In this model, we added a measure of the temperature used to elicit WU as a covariate. In particular, we entered the actual temperature minus the average temperature used in successful windup trials (49.5° C). As a centered measure, the value of the intercept in this model and in model 4 (below) remains interpretable, as the initial NPS report was in Models 1 and 2. Addition of this variable significantly reduced unexplained variability at level 2 by about 10% (pseudo- R²). Inspection of the regression coefficient indicates slightly more than a 1:1 increase in the first report of aftersensations for each degree increase in temperature. Thus, there were significant differences between people in the temperature used to elicit WU, and higher temperatures tended to produce greater reports of aftersensations at the first post-stimulus interval.

Model 4, add group and site variables

The next model added group, site and their interactions as covariates. Analysis revealed a significant interaction between patient status and site in rate of change. Figure 2 shows reports of aftersensations over time based on the model’s predictions, as a function of patient status and site. Inspection shows a shallower slope when patients were tested in the face (upper right), indicating that the decay of aftersensations was slower for patients when tested in the area of their clinical pain. Computation of individual regression coefficients for each combination of patients group and site showed the shallowest slope (95% confidence limits) over time for the patient-face condition [−1.21 (−1.06, −1.36)]. The slope for the patient hand condition was −1.97 (−1.87, −2.07) and the respective values for the control subjects were −2.38 (−2.19, −2.58) and −1.81 (−1.57, −2.04). Relative to Model 3, pseudo-R² indicated a 4.3% reduction in level 2 variability (between persons variability in initial status) attributable to the grouping variables, and an 8.1% reduction in level 1 variability (within person variability in true trajectory of change). Neither model 3 nor 4 suggested that initial status was correlated with rate of change. Finally, the presence of non-zero estimates of unexplained variability in Model 4 at both level 1 and level 2 suggests that future research aimed at identifying time varying covariates, as well as additional grouping factors, that might reduce, respectively, these unexplained sources of variation is warranted.

Figure 2.

Model-based numeric pain scale ratings, adjusted for temperature, broken down by anatomic site and group (C: Control, P: Patient)

Exploratory analysis of changing WU criteria

As a last analysis, we explored the effect of changing criteria for the definition of WU on this analysis. First, reducing the necessary increase from 25 to 20 points (but keeping the requirement that the response to the first stimulus be at or below the minimum pain level) had little effect on the statistical results. In particular, analysis of aftersensations using those data, which added 3 subjects and almost 25% more WU trials, showed a 3-way interaction of group, site and time (p= .02) and a level of unexplained variance (32.1) that were consistent with the analysis presented in Table 3. Second, however, analysis of aftersensations provided by allowing any 25-point increase in ratings, which added a similar number of WU trials to the analysis, failed to show the 3-way interaction (p= .40) and more than doubled the amount of unexplained variance (67.0). Thus, these analyses suggest that while there was little effect of reducing the necessary increase in report level from 25 to 20 points, there was a strong effect of requiring that initial report levels be non-painful.

Summary of main analyses

To summarize our main analyses, Models 1 and 2 showed, respectively, that there was variability in average report levels, and in decay of aftersensations that could potentially be related to variation in patient and site. In addition, Model 1 showed that the correlation among the residuals made repeated measures ANOVA an inappropriate model for analyzing these data. Model 3 showed that the level of aftersensations varied directly with the temperature used to elicit wind-up, and that it was valuable to control for these differences in further analysis. Finally, Model 4 showed that the rate of decline in aftersensations was slowest in patients when tested on the face. With regard to the a priori questions, the data suggest that WU may be elicited on the face, and while there were no main effects of either site or group on the decay of aftersensations, aftersensations did decay more slowly in the patients, but only when WU was elicited from the skin overlying the masseter muscles.

Discussion

This initial investigation of trigeminal versus extratrigeminal WU in TMD patients and controls showed that it is possible to elicit thermal WU in both the hand and face. It was more likely that WU would be elicited from hand than facial stimulation, but similarly likely to be elicited from patients and controls. For both anatomic locations, this study failed to demonstrate significant differences in the slope of WU for myofascial TMD cases versus control, unlike two prior investigations.^{4 31}. However, similar to earlier reports by Maixner and colleagues⁴ and Sarlani and colleagues³¹, TMD patients were likely to reach WU asymptote sooner than controls. Most important, this study showed that decay of painful aftersensations was significantly slower for myofascial TMD patients, compared to controls, when tested in the trigeminal region (i.e., skin overlying the body of the masseter muscle), but not in an extratrigeminal region (i.e. thenar eminence). Thus, these results suggest that enhanced central sensitivity in TMD may be largely confined to the region of clinical pain and not widespread, as generally observed in comorbid conditions such as fibromyalgia.

One possible difference between this study and the two prior studies that found differences in trajectory of WU between TMD cases and controls^{4 31} is that we did not exclude controls with other pain conditions. Both prior studies explicitly selected “pain free” controls. By violating assumptions on which case/control studies are based (i.e., controls representing those who would have been selected as cases, had they developed the disease being studied; and use of the same exclusion criteria for cases and controls (see^{43 36}) a bias toward finding group differences is likely in these earlier studies, because enhanced general pain perception may be associated with multiple conditions that are allowed in the TMD case group (e.g., fibromyalgia, low back pain) but stripped from the healthy control group. Notably, three studies of experimental pain in TMD patients that correctly selected controls solely on the basis of absence of orofacial pain^{5, 14, 15} did not find significant case-control differences. Our failure to find pain threshold differences between cases and controls in either the trigeminal or extratrigeminal region is, therefore, consistent with other experimental pain studies that employed similar strategies for recruitment of controls. Thus, differences between TMD patients and TMD-free controls in experimental pain response may be even less robust than suggested by some of the earlier literature that most often recruited pain-free controls.

These factors may have led to a conservative group comparison in relation to prior studies on WU in TMD. Thus, it is particularly impressive that, despite the conservative design, we documented unique differences in decay of aftersensations in the trigeminal region among TMD patients versus controls, despite failure to find case-control differences in decay of aftersensations in an extratrigeminal region. The finding of prolonged aftersensations specific to the trigeminal region in TMD patients deserves further attention. If prolonged aftersensations are confined to the region of clinical pain in TMD patients, it suggests that central synaptic sensitivity is primarily influenced by nociceptive signaling from the temporomandibular joint and/or masticatory muscles and is not a phenomenon of global enhancement of central nervous system.

One might also question whether this study failed to find differences in WU slopes between groups or body sites due to its restrictive data inclusion process. In analyses not detailed above, we conducted an analysis that included all trials attempting to elicit a WU response which produced at least one report of aftersensations. The effect of this inclusion was to raise the average first report to about NPS= 25 and the 15^th report to only about NPS= 35, greatly diluting the magnitude of temporal summation and increasing the magnitude of error variance. Most importantly, these data still did not show any difference between patients and controls in the slope of WU. Thus, the failure to show robust differences in WU between patients and controls does not appear to be the result of the strict data inclusion process.

One possible criticism of our study is that TMD patients’ prolonged painful aftersensations in the trigeminal region might have been a function of deep tissue stimulation rather than WU. Most subjects reported that the probe produced a ‘tap’ sensation when energized, particularly on the face. However, the shallow contact seems unlikely to have caused deep tissue stimulation. Nevertheless, to firmly rule out this explanation, it would have been useful to test the response to repetitive contacts of the face at ambient temperatures, using the same stimulus delivery mechanism employed here. It will be important to replicate the current findings, using newer technology ⁴⁴ that avoids the need to cycle thermode contact against the skin.

Another possible criticism of our methods is that allowing subjects to hold the thermode assembly themselves may have introduced variation in contact among subjects, introducing error or bias. Note, however, that participants were not holding the thermode but the thermode assembly, with instructions to place their hands or face on the plexiglass apparatus. Thus, it seems unlikely that there were large differences among participants in the amount of skin stimulated or the amount of force against the thermode. Moreover, even if there were variation in thermode assembly placement, it is hard to imagine how this could account for the very specific finding of group differences in patterns of aftersensations restricted to the face, when the thermode had been removed from the skin.

It is notable that not all patients or controls provided subjective pain ratings that met our strict definition of WU. In fact, although the proportion of cases and controls providing WU data (i.e., 53% versus 71%) was not significantly different, these data demonstrate a potential difficulty in achieving a subjective WU response in human subjects, when defining WU according to strict criteria (see Methods). We suspect that it may have been somewhat more difficult for a pattern of NPS ratings to meet WU criteria when stimulating the face than the hand, because the decreased distance to the CNS when stimulating the face makes it more difficult for subjects to discriminate first from second pain sensations. In results not detailed above, we were unable to identify any clinical or demographic characteristics that differentiated individuals who were able to produce a WU response from those who were not. Nevertheless, because of the failure to generate a WU response in many participants, caution is warranted in interpreting these findings. Replication is essential.

To our knowledge, this is the first study that explicitly documents problems in achieving a robust WU response in the large majority of subjects. Unlike prior studies in which much higher proportions of subjects are reported to achieve WU criteria (e.g., ⁴⁵: 55 of 58 subjects or nearly 95%), a notably smaller proportion of subjects in the current study met WU criteria. Given the discrepancy in the proportion of subjects achieving WU in this study versus prior published literature, we considered possible sources of the discrepancy.

Although we trained subjects during pre-study practice trials to differentiate between pain and sensitivity due to pressure contact, the differentiation may have still been difficult for some subjects. Moreover, we spent a maximum of two hours with each subject; some prior studies (e.g., ⁴⁶) have likely spent considerably more time on subject training, including the use of multiple training sessions. In addition, the anatomic sites used, i.e., the thenar eminence and the origin of the masseter muscle, have limited area. In searching for an adequate WU stimulus, if a too-high stimulus intensity was tried (i.e., one that elicited a response greater than NPS= 20), this limited surface area may have fostered the development of peripheral sensitization that interfered with later elicitation of a WU response This problem could be avoided in the future by using various areas of the hands or arm, for example. In any case, improvement in technology, using a device that can elicit heat pulses with continuous thermode contact ⁴⁴ may prove to be a more efficient mechanism for eliciting WU, particularly in the trigeminal region.

In sum, these novel findings demonstrate the feasibility of generating a WU response in the trigeminal region, in both TMD patients and non-TMD controls. The finding of prolonged painful aftersensations in TMD patients, specific to the trigeminal region, deserves replication in larger studies and with newer methods for generating a robust WU response⁴⁴. If replicated, additional studies designed to better understand the mechanisms underlying prolonged aftersensations specific to the trigeminal region are warranted.