1 / 24

What is TMD? Connections to Chronic Pain

What is TMD? Connections to Chronic Pain. Temporomandicular Disorder (TMD) is a chronic pain disorder defined as extended pain in the orofacial region, particularly in the masticatory muscles or at least one temporomandibular joint.

doli
Download Presentation

What is TMD? Connections to Chronic Pain

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. What is TMD? Connections to Chronic Pain • Temporomandicular Disorder (TMD) is a chronic pain disorder defined as extended pain in the orofacial region, particularly in the masticatory muscles or at least one temporomandibular joint. • Ranks second only to headache as the most likely clinical condition to cause craniofacial pain. • Patient treatment is largely restricted to symptomatic care • Other chronic pain disorders include migraine headaches and fibromyalgia, which are comorbid with TMD • Though this study was restricted to studying TMD, our study is relevant to all chronic pain disorders. By constructing a model to identify risk factors of TMD, we hope to identify the biological mechanisms controlling the body’s response to painful stimuli. Such information would ultimately change the way chronic pain is studied and treated.

  2. Orofacial Pain: Prospective Evaluation and Risk Assessment (OPPERA) Study • Purpose: Establish the causal determinants of TMD pain. • Prospective cohort design of 3200 people who did not have TMD when enrolled as well as a case-control study that enrolled 200 TMD cases • Based on a model with enhanced pain sensitivity as an intermediate phenotypic risk factor for TMD development • Baseline assessment of pain sensitivity from quantitative sensory testing (QST), including thermal pain sensitivity measures

  3. Thermal Pain Sensitivity Testing-Protocol • Heightened sensitivity to pain (including thermal pain) is believed to be a risk factor for TMD • Subjects were given a series of heat pulses with a thermode (a small metal block) and asked to report a pain rating between 0 and 100 • 3 trials with peak temperatures of 46°C, 48°C, 50°C • 10 pulses for each stimuli, pulses applied at the same temperature successively • If subject rated their pain as 100, they were given the option to continue or end these tests. • Following the 10th pulse, the subject gave lingering pain rating at intervals of 15 and 30 seconds (referred to as aftersensations)

  4. Temporal Summation and Chronic Pain • There are two risk factors for pain: overall sensitivity (measured by the first pulse) and temporal summation • The primary reason for the 10 consecutive pulses was to assess temporal summation, sometimes referred to as “windup”. • It is believed there is a difference between cases and controls in the amount of the subjects will increase their pain rating when exposed to repetitive painful stimuli. Subjects with a higher first pulse (overall sensitivity) are also more likely to be cases. • Though a common theory in the pain world, temporal summation as a predictor of chronic pain disorders has been largely unstudied. • This study indicates an optimal predictor of case status combines a measure of general sensitivity to pain and a measure of temporal summation. • These results are intriguing for chronic pain specialists, since temporal summation is believed to be partly responsible for the transition from acute to chronic pain.

  5. Mean Pain Ratings Over Time --- Control __ Case There is clearly a difference between the patterns of the cases and controls in this windup data. Our goal is to understand the mechanisms causing these differences from a quantitative standpoint.

  6. Existing Methods for Analyzing Windup Data • First Pulse - Subject’s pain rating at time • Higher the first pulse, the higher the risk of TMD • Relatively weak association • Does not consider an increase or decrease in pain after the first pulse, so does not provide any information on temporal summation • Area Under the Curve- Plot the pain rating against time, taking the area underneath the curve • More strongly associated with case status • Thought to better account for any windup since it uses all 10 ratings • Highly correlated with the first pulse statistic (correlations of at least 0.81 at each temperature)

  7. Existing Methods for Analyzing Windup Data • Delta- Difference between the maximum pain rating over all ten pulses and the first pulse. • Only weakly associated with case status • Problem: “ceiling effect” • For subjects whose first pulse is very high, it is difficult to obtain a large increase since the maximum possible rating is 100 • Since subjects whose first pulse is very high are more likely to be cases, this measure was a poor predictor of case status.

  8. Existing Methods for Analyzing Windup Data • Regression Slope- Slope of the regression line that predicts the pain rating based on time for the first three ratings. (both with or without an intercept) • With intercept: not significantly associated with case status • Problem of ceiling effect: Subjects with high first pulse, have little room to increase before reaching the maximum rating, giving regression slope of nearly 0 • Highly correlated with delta measure (correlations of 0.8 at each measure) • Without intercept: associated with case status • Highly correlated with the first pulse measure and area under the curve (correlations of 0.94 at each temperature)

  9. Additional Predictors • Maximum • Maximum pain rating for each temperature • Strongly associated with case status at every temperature • Again, correlated with the first pulse rating and the area under the curve (correlation of 0.7 or higher) • Aftersensation • 2 measurements from the raw data • 15-second aftersensation rating at 50 degrees was the strongest predictor observed (compared to classical variables and raw data)

  10. Lasso Models • Penalized regression used to determine the optimal combination of variables for a multivariate regression model for predicting case status • Useful when the predictors in a regression model are correlated with one another • Penalizes models with too many nonzero coefficients by minimizing the following: xij‘s are the predictors yi is the response variable bj‘s are the LASSO regression coefficients λ is a tuning parameter

  11. Lambda and Cross-Validation • λ is a tuning parameter that specifies how much each coefficient is penalized • To choose our lambda, we use cross-validation • Data set is split into ten approximately equally-sized partitions. • For each partition, we fit a model using the 90% of the data that is not in the partition. • Attempt to predict the values of y for the 10% that is in the partition. • Repeat this procedure for each partition and for various λ values. • Use our predicted values of y to calculate the area under the receiver operating characteristic (ROC) curve (denoted by AUC) in each left out partition.

  12. Cross-Validation Curve with Classical Variables, Raw data, and Aftersensation This is our cross-validation curve with our raw data values (30 pulses), the derived measures described previously, and the aftersensation ratings. The numbers on the top indicate the approximate number of variables in the model. The minimum lambda gives a model with 2 variables.

  13. Lasso Approach 1All derived variables as well as raw data Best AUC Surviving Variables: Site ID, Max Rating, 46°

  14. Lasso Approach 2All derived variables, Raw data, Aftersensations Best AUC Surviving Variables: Site ID, Aftersensation at 15 s, 50°

  15. Notes on Lasso • Tempting to speculate that the maximum measurement and aftersensation are more informative than all of the other measures. • However, if there are several equally good predictors that are correlated with one another, LASSO tends to arbitrarily pick one of them, even if an alternative predictor would produce comparable accuracy. • However, LASSO shows that if we know that if we know one variable (in this case maximum and/or aftersensation), we know as much about case status as we would if we included all the variables

  16. First Pulse and Windup Relationship • As discussed earlier, overall sensitivity to pain (measured by first pulse) is believed to be distinct from temporal summation (measured by delta) • Maximum rating = First pulse rating + Delta rating • Interesting to note that the strongest predictor of cases status (when aftersensations are excluded) is the sum of these two measures • To better understand the relationship between first pulse and delta, the subjects were separated into ten groups based on first pulse (0-9, 10-19, etc).

  17. Delta Grouped by First Pulse The plots indicate that subjects with very low first pulse generally had very low delta values as well. Subjects in the 20-60 range had much higher deltas. Finally, subjects with very high first pulses had low delta values. One might expect this last group had low delta values as a result of the ceiling effect. However, this is clearly not the case since the medians in each group are well below 100.

  18. Bootstrap Analysis • To strengthen this analysis, it was necessary to compare the delta values in each first pulse grouping. • Typically, one would take the mean of the three groups and test the null hypothesis that they are all equal. • However, this comparison was problematic because the mean delta values in each first pulse grouping could be restricted by the ceiling effect. • There may be some people in the first pulse group over 60 who stopped at the 100 rating who actually would have continued to an even higher rating. • The true mean of the deltas of this first pulse group might behigher than the mean we actually calculate. • Solution: use the median. As long as the median is less than 100, it is unaffected by this ceiling effect.

  19. Bootstrap Results • In order to compare medians, we need to use bootstrap analysis, as out statistical distribution is unknown. • We created 3 bands in each trial based on first pulse ratings (0-19, 20-59, 60-99). • We then used bootstrapping to find confidence intervals for the median delta in each band and to test the hypothesis that the medians are the same in all the bands. • Null hypothesis: the median deltas in all the bands were equal. • Found that the middle band had significantly higher median deltas than the lowest and highest bands (bootstrap confidence intervals never overlap)

  20. Confidence Intervals from Bootstrap Analysis At each temperature, the 95% confidence intervals of groups 1 and 3 never overlap with group 2. Clearly there is a difference between the median deltas in these groups.

  21. Mean Pain Ratings Over Time There are distinct different patterns occurring between first pulse groups, temperatures, cases, and controls. While we have found several strong predictors of TMD case status, the story of temporal summation is not nearly complete. Temporal summation is only a predictor of case status for a certain subset or the population.

  22. Number of Members in Each Subgroup

  23. Delta as a Predictor in Each Subgroup 1-Standardized Odds Ratio (SOR)- the predictor was transformed to a unit-normal deviate prior to fitting the logistic regression model. This transformation means that odds ratios can be interpreted as the relative increase in odds of TMD for each standard deviation increase in the variable.

  24. Conclusions • Our results suggest possible improvements for the thermal testing protocol • It may not be necessary to repeat the experiment three times at different temperatures, since the results do not vary greatly with respect to temperature • Higher temperatures may be better when measuring aftersensations • Our results also increase our understanding of chronic pain • Overall thermal sensitivity and temporal summation are related, but the relationship is complex • Better understanding of temporal summation may lead to novel methods to treat and prevent chronic pain

More Related