The Longitudinal Relationship Between Emotion Regulation and Pain-Related Outcomes: Results From a Large, Online Prospective Study.

These data were part of a large, online, prospective study investigating relationships between pain and emotional experiences in adults with chronic pain recruited in the community. Data were collected at baseline (Time 1 [T1]; April – May, 2020) and at three-month follow-up (Time 2 [T2]; July – August, 2020). This study was not preregistered. The analyses reported in this article reflect primary study aims and are distinct from prior publications. One secondary data analysis paper has been published from this study, examining the early impact of COVID-19 on chronic pain, based on T1 data only. 57 This study was approved by the Johns Hopkins University (JHU) Institutional Review Board (IRB # IRB00224112). In accordance with the JHU IRB policy for exempt applications, whereby the identity of human subjects cannot be readily ascertained, each participant was provided with an overview of the study, the contact information of the principle investigator, and the JHU IRB number prior to agreeing to participate.

Table 1 summarizes the socio-demographic and baseline pain characteristics of the study sample. The breakdown of gender, race, and ethnicity in the present sample was similar to that of the nationally representative online chronic pain population reported by Johannes and colleagues 40 . The mean age of the participants was 41.7 years (SD = 13.1), and the participants were predominantly female, White, had at least some college education, and working either part- or full-time. The majority of participants had an annual income above $24,999 and were married. On average, participants reported 10.9 years (SD = 9.1 years) of experiencing chronic pain. Participants’ baseline pain severity levels based upon BPI on average was 4.1 (SD = 1.6). This is below average compared to treatment seeking chronic pain samples (e.g., 6.98). 79 On the WPI, participants endorsed pain in an average of 5.3 (SD = 3.5) out of 19 possible bodily regions on the WPI. The primary pain condition experienced by participants was low back or neck pain (65.4%). Other conditions endorsed included migraine (33.3%), arthritis (20.7%), irritable bowel syndrome (17%), fibromyalgia (11.5%), temporomandibular disorder (4.6%), and endometriosis (3.9%). A total of 39.6% (575 out of 1453) participants’ data were missing at the three-month follow-up assessment. We conducted a series of attrition analyses using t-tests (for continuous variables) and chi-square tests (for categorical variables) to examine if there were certain groups of participants who were more likely to drop out in the 3-month follow-up assessment. We found that participants who were younger ( p < .001), Hispanic ( p < .001), or had lower education level ( p < .01) were more likely to drop out in the follow-up assessment. Neither the duration of chronic pain ( p = .2) nor pain severity at baseline ( p = .06) were associated with the missingness. The results of our attrition analyses lend further support that the missingness in the present study may meet the MAR assumption, which means that there is an association between the missing values and some of the observed data. Table 2 presents means, standard deviations, and bi-variate correlations among key study variables. The correlations among study variables indicated relationships that, at their strongest, were moderately correlated (see Table 2 ). Tolerance ranged from .52 to .91 and VIF ranged from 1.10 to 2.05. Hence, there was no evidence for multicollinearity among key study variables. The model fit was not provided, as the current model is fully saturated. Table 3 provides the detailed path estimates of the model and main findings from Model 1 are depicted visually in Figure 1 . Controlling for socio-demographic covariates and baseline pain severity and pain interference, pain acceptance ( B = -.102, SE = .028, p = .001) at baseline was significantly associated with pain severity at 3-month follow-up. In other words, individuals with greater pain acceptance at baseline reported lower pain severity at three-month follow-up. Controlling for socio-demographic covariates and baseline pain severity and pain interference, pain catastrophizing ( B = .017, SE = .006, p = .006) and pain acceptance ( B = -.054, SE = .009, p < .001) at baseline were significantly associated with pain interference at three-month follow-up. Of the socio-demographic covariates, none were significantly associated with pain severity or pain interference at three-month follow-up. The model fit was not provided, as the current model is fully saturated. Tables 4 shows the detailed path estimates of this model and the main findings of Model 2 are depicted visually in Figure 1 . Controlling for socio-demographic covariates, baseline pain severity, pain interference, and pain-specific emotion regulation, alexithymia (but not other measures of general emotion regulation) at baseline was significantly associated with pain severity at three-month follow-up ( B = .012, SE = .005, p = .032), such that individuals with higher levels of alexithymia at baseline reported greater pain severity at 3-month follow-up. Among pain-specific emotion regulation variables, pain acceptance at baseline remained significantly associated with pain severity at three-month follow-up ( B = -.022, SE = .007, p = .001). Controlling for socio-demographic covariates, baseline pain severity, pain interference, and pain-specific emotion regulation, alexithymia (but not other measures of general emotion regulation) at baseline was significantly associated with pain interference at three-month followup ( B = .026, SE = .007, p < .001). Pain acceptance at baseline remained significantly associated with pain interference at three-month follow-up ( B = -.053, SE = .009, p < .001); however, the association between pain catastrophizing at baseline and pain interference at three-month follow-up became statistically non-significant ( B = .009, SE = .007, p = .153). Of the socio-demographic covariates, age significantly predicted both pain severity ( B = .007, SE = .004, p = .042) and pain interference ( B = .012, SE = .005, p = .017), such that older age was associated with greater pain severity and interference at three-month follow-up. First, we conducted a path model that included only baseline pain severity, pain interference, and socio-demographic covariates. Inclusion of these covariates explained 46.9% and 45.4% of variance of pain severity and pain interference measured at 3-month follow-up, respectively. Model 1, which included pain-specific emotion regulation variables in addition to these covariates, accounted for additional 0.6% and 2.6% of variance in pain severity and interference, respectively. Finally, Model 2, which included general emotion regulation variables and pain-specific emotion regulation variables, accounted for additional 0.3% and 0.8% of variance in pain severity and interference, respectively. Next, because pain severity and pain interference at baseline accounted for much of the variance in pain severity and pain interference at follow-up, we isolated variance accounted for by pain-specific and general emotion regulation variables in a second series of effect size models, excluding baseline pain severity and pain interference. In these models, pain-specific emotion regulation variables at baseline accounted for 15.8% and 30.2% variance for pain severity and interference at 3-month follow-up, respectively. General emotion regulation variables at baseline accounted for 4.5% and 9.5% variance for pain severity and interference at 3-month follow-up, respectively. Due to methodological concerns explained in the Methods, we examined statistical models using two select TAS-20 subscales (DDF, DIF; TAS-12) and excluding a third (EOT). However, chronic pain studies more typically examine the full TAS-20 measure. Thus, to ensure the robustness of models presented and to facilitate comparison with the extant literature examining alexithymia in chronic pain, we replicated our statistical models with the full TAS-20 score. This model was statistically similar, and alexithymia remained a statistically significant predictor of pain severity ( B = .012, SE = .005, p = .032) and interference ( B = .026, SE = .007, p < .001). Next, to address the potential concern that the DIF subscale of the TAS-20 has sensory items that might drive significant associations with pain-related outcomes, we conducted another sensitivity analysis that compares results when only including DIF vs. DDF subscales of the TAS-20 in the model. We found that while controlling for all covariates, DIF was significantly associated with both pain severity ( B = .021, SE = .008, p = .012) and pain interference ( B = .043, SE = .011, p < .001) at 3-month follow-up. On the other hand, DDF was no longer significantly associated with pain severity ( B = .016, SE = .013, p = .219), but was significantly associated with pain interference ( B = .045, SE = .018, p = .011).

The current study determined the extent to which general difficulties with emotion regulation were associated with pain severity and pain interference three months later, above and beyond pain-specific emotion regulation, in a large sample of adults with mixed chronic pain conditions recruited from the community. Using conservative models that accounted for baseline pain severity, pain interference, and socio-demographic factors, pain specific and general emotion regulation only accounted for a small portion of the variance in pain severity and pain interference. However, with indices of general emotion regulation added to the model, statistically significant relationships between pain catastrophizing and pain interference became nonsignificant. Instead, greater alexithymia—difficulty identifying and describing one’s emotions—emerged as a statistically significant predictor of pain-related outcomes. Findings suggest that alexithymia may be valuable to assess alongside pain-specific emotion regulation strategies in order to better understand pain-related outcomes. Emotion regulation involves identifying one’s emotions and engaging in strategies to maintain or alter one’s emotional state. 32 , 33 Alexithymia reflects difficulties with the first stage of emotion regulation, emotion identification. 33 , 42 People with (versus without) chronic pain report higher levels of alexithymia, 29 , 41 , 66 , 90 and alexithymia is related to higher pain severity and pain interference 21 , 36 and greater depression and anxiety symptoms. 21 , 50 , 66 Consistent with the present study, prospective studies have found that alexithymia predicts poor pain-related outcomes. 10 , 68 The current study extends prior work by demonstrating that significant relationships between pain catastrophizing and pain interference became nonsignificant after accounting for alexithymia and other aspects of general emotion regulation. This was somewhat surprising, given robust associations between pain catastrophizing and pain interference in the literature. Several possibilities warrant future research. Difficulty with emotion identification can have cascading effects that impair one’s ability to select and implement strategies to successfully alter emotional trajectories. 33 , 42 , 77 Elevated alexithymia may capture a broader and more robust emotion regulation difficulty that contributes to difficulty functioning despite chronic pain. Pain demands attention, particularly when the threat value associated with pain is high. 22 Hypervigilance towards pain may detract attention away from relevant affective cues hindering emotional awareness and contributing to poor pain-related outcomes. Pain acceptance remained significantly associated with pain-related outcomes after accounting for general emotion regulation. The measurement approach of this study might have influenced this finding. We focused on well-established measures of pain-specific emotion robustly related to pain-related outcomes, and, general emotion regulation processes that align with recent affective theory (i.e., 33 ). We prioritized widely used questionnaires with precedence for use in chronic pain. As a first step, use of a well-validated and theory-driven battery helps increase confidence in study findings. However, this approach yielded some discordance in the measurement of pain-specific and general emotion regulation. For example, the alexithymia questionnaire assessed awareness of interoceptive cues related to emotion but there was no complimentary assessment of interoceptive cues related to pain. Results from a sensitivity analysis suggested the relationship between alexithymia and pain severity was driven in part by sensory focused items. The chronic pain acceptance questionnaire assessed acceptance of chronic pain, but there was no complimentary assessment of general acceptance. One cross-sectional study found that pain acceptance and general acceptance each accounted for unique variance in mood and pain outcomes. 54 suggesting a similar pattern might have emerged in the current study. A next step in this line of work is to develop and test more concordant models of assessment. General emotion regulation strategy engagement has been infrequently studied in chronic pain. 45 In the current study, cognitive reappraisal (reinterpreting an emotionally evocative situation) and expressive suppression (attempting to inhibit emotional expression) were significantly related to pain interference at baseline but not three-month follow-up. Emotional approach coping (via processing and expressing emotions) was not related to pain severity or pain interference. This was unexpected, given conceptual overlap between pain-specific and general emotion regulation; however, some other studies have also reported nonsignificant relationships between general emotion regulation strategies and pain. 82 , 88 Difficulties with emotion identification may pose global emotion regulation difficulties, whereas difficulties with strategy engagement may be more dependent on contextual and individual factors in people with chronic pain. One study demonstrated a negative relationship between emotional approach coping and pain intensity and pain interference; however, these relationships were mediated by negative affect and moderated by age and sex. 91 General emotion regulation strategies may be more relevant to pain outcomes among individuals with co-occurring psychopathology or substance use. For example, cognitive reappraisal is less common and less successful among individuals with chronic pain who misuse prescription opioids compared to those who do not. 27 Finally, emotion regulation strategies have historically been viewed as “adaptive” or “maladaptive” (e.g., with catastrophizing viewed as maladaptive and reappraisal viewed as adaptive); however, emerging research suggests that emotion regulation is more successful when an individual utilizes a mix of diverse strategies. 4 , 11 More research is needed to clarify the role of emotion regulation in chronic pain. Latent profile analysis may be helpful in characterizing how constellations of pain-specific and general emotion regulation are associated with pain related outcomes. The findings of the current study have implications for intervention. Chronic pain interventions often emphasize pain-specific emotion regulation. For example, reframing catastrophic thoughts about pain is a cornerstone of Cognitive Behavioral Therapy (CBT) for chronic pain, and cultivating pain acceptance is central to Acceptance and Commitment Therapy for chronic pain. 53 Current results suggest potential value in expanding chronic pain intervention to target emotion regulation more broadly, particularly identifying and describing emotions. Emotional Awareness and Expression Therapy (EAET) was recently developed to help people with chronic pain identify, experience, and express difficult emotions. 49 Compared to CBT, EAET has been shown to result in greater improvements in pain intensity 49 , 89 and widespread pain. 89 Another emotion-focused exposure intervention, the Hybrid Treatment, targets pain-specific and general emotion regulation strategy engagement; compared to CBT, the Hybrid Treatment resulted in greater improvements in mood and pain interference among individuals with chronic pain and elevated depression and anxiety symptoms. 12 For some, treatment targeting both pain-specific and general emotion regulation strategies may be more effective for improving pain-related outcomes. This study has several notable strengths. First, the current sample was comprised of a large online sample of adults with chronic pain across the United States. Second, we recruited a demographically representative sample of people with chronic pain in the United States. Third, the statistical models presented in this paper are robust, accounting for baseline pain severity, pain interference and sociodemographic covariates. This study has several limitations. We relied exclusively on self-reported measures, which are limited by response biases and retrospective recall. Although the TAS-20 is the most common assessment of alexithymia in chronic pain, 2 the EOT subscale has been criticized for poor psychometric performance and confounding cultural norms of emotional expression. 20 , 46 We addressed this in part by excluding EOT from analyses and conducting sensitivity analyses; however, more research is needed to improve the assessment of alexithymia in chronic pain. Future studies are needed to expand methodological assessment of emotion regulation in chronic pain; for example, laboratory-based paradigms are available to assess emotion identification 3 and emotion regulation strategy engagement in vivo. 31 Beyond self-report, various psychophysiological and neuroimaging methods can be used to obtain useful objective indices of emotion regulation. 27 , 31 Emotion regulation is dynamic, and future studies are needed to assess emotion regulation using more intensive longitudinal sampling to examine real time changes in emotion regulation. A few additional methodological notes are warranted. Oversampling was effective for achieving a representative sample; however, our sample was still predominantly White (82%) and Non-Hispanic (89%). Oversampling to improve representation beyond the chronic pain population could have been more inclusive and provided greater power to examine demographic covariates and moderators. Results should be considered in the context of the data collection timeline, which began shortly after the beginning of the COVID-19 pandemic and implementation of social distancing in the United States. Initial data suggests that COVID-19 may influence psychosocial and pain-related variables. 37 We previously reported the perceived impact of COVID-19 on participants in this study at Time 1; the majority reported no or minimal change in pain intensity and pain interference as a result of COVID-19. 57 Nevertheless, changes in pain coping or engagement in pain intervention related to COVID-19 or otherwise might have occurred over the course of data collection. The current study tested the hypothesis that general emotion regulation predicts pain-related outcomes, over and above pain-specific emotion regulation. We found that difficulties with emotion regulation at baseline, particularly alexithymia, was associated with greater pain severity and pain interference at three-month follow-up. Moreover, after accounting for general emotion regulation, pain catastrophizing was no longer a statistically significant predictor of pain interference. When considering emotional factors that impact pain outcomes, the current study highlights the value of considering general emotion regulation in addition to pain-specific constructs. More research is needed to elucidate the relationship between emotion regulation and pain-related outcomes, as well as individual difference factors that moderate these relationships. Ultimately, this line of research may shed light on risk for poor pain outcomes and inform further refinement of psychological interventions for chronic pain.

Pain is a subjective experience characterized by unpleasant sensory and emotional qualities 70 . Similar to pain, emotions are subjective, and are influenced by one’s interpretation of interoceptive cues, situational context, and past experiences. 8 , 9 , 30 Emotion and pain are generally adaptive, but can become maladaptive in the absence of a threat, 30 , 61 , 70 and thus often require attention and effort to regulate. Research consistently shows that an individual’s approach to managing the emotional aspects of pain, especially via pain catastrophizing and pain acceptance, reliably predicts future pain-related outcomes. 67 , 69 , 80 However, it is unclear the extent to which an individual’s approach to managing emotions more broadly relates to pain-related outcomes. Emotion regulation is the multifaceted process of altering one’s emotional trajectory towards a desired state. Recent affective science theory states that emotion regulation occurs by first identifying one’s emotional state and next engaging in strategies to up-regulate (e.g., enhance) or down-regulate (e.g., attenuate) emotion. This process is used to modulate both positive (e.g., gratitude) and negative (e.g., anger) emotions. 33 , 34 Emotion regulation can occur with or without conscience effort, although explicit emotion regulation (i.e., occurring with conscious effort) has received more empirical attention and is often emphasized in psychotherapeutic interventions. 13 , 81 In our theoretical review on emotion regulation in pain, we describe how processes such as pain catastrophizing and pain acceptance can be viewed as strategies to regulate emotions specific to the experience of pain. 1 Pain catastrophizing, an exaggerated negative cognitive orientation towards pain, 69 often emerges as an attempt to down-regulate fear of pain. For example, pain catastrophizing can be used to elicit support from others during periods of distress 76 or to down-regulate distress via problem solving. 23 , 62 Ultimately, however, pain catastrophizing tends to up-regulate fear of pain and increase the experience of pain itself. 84 , 85 Pain acceptance—a tendency to engage in valued activities despite pain 51 —can be employed to down-regulate fear of pain by acknowledging uncomfortable emotions and engaging in goal-directed action despite them. 51 , 83 Greater use of pain catastrophizing relates to the development, persistence, and severity of chronic pain. 67 , 69 , 80 Chronic pain acceptance, on the other hand, is associated with lower pain intensity, 39 , 51 improved pain-related functioning 39 , 59 and improved psychological wellbeing. 51 , 52 , 83 Often, pain catastrophizing is considered “maladaptive,” and pain acceptance “adaptive,” for managing negative pain-related emotions. Although, it is important to note that the “success” of emotion regulation depends on factors such as culture, developmental stage, and situation. 4 , 56 , 63 Difficulties regulating the emotional aspects of pain may reflect broader emotion regulation difficulties. 54 For example, difficulties reappraising catastrophic thoughts about pain may relate to difficulties reappraising catastrophic thoughts more broadly . General difficulties regulating emotions have been identified as potential transdiagnostic factors underlying chronic pain and common chronic pain comorbidities, such as psychopathology and problematic substance use. 1 , 45 , 47 Thus, if general emotion regulation helps explain pain-related outcomes over and above pain-specific emotion regulation, it may represent a broader treatment target that can improve both pain- and mood-related outcomes relevant to people with chronic pain. The current study recruited a large community sample of adults with chronic pain to determine the extent to which general emotion regulation (i.e., emotion identification and strategy engagement) at baseline was associated with pain-related outcomes (pain severity, pain interference) at three-month follow-up. Based on theoretical overlap, we hypothesized that general emotion regulation would account for unique variance in pain-related outcomes over and above pain-specific emotion regulation (i.e., pain catastrophizing and pain acceptance). Within this general model, we did not make hypotheses about specific indices of general or pain-specific emotion regulation.

Participants were recruited through Amazon Mechanical Turk (“MTurk”), an online platform that connects hundreds of thousands of MTurk “workers” to behavioral health researchers. 48 Compared to traditional, in-person, research methods, MTurk provides access to large and diverse samples across the United States. Mounting evidence demonstrates that data collected from MTurk workers have good validity and reliability, comparable or superior to in-person data collection methods. 14 , 17 MTurk can provide unique access to individuals with specialty health conditions and has been used to recruit large samples of people with chronic pain. 6 , 14 In the current study, we worked with a third-party platform, Cloud Research, to recruit adults with chronic pain and to achieve a baseline sample demographically representative of people with chronic pain in the United States. 40 Initial pilot testing suggested that people with chronic pain who identified as Black or African American, Hispanic, and women were under-represented in the MTurk sample; thus, these individuals were oversampled to better reflect the chronic pain population in the United States. 40 Inclusion criteria for this study included: (1) age 18 or greater; (2) the presence of chronic pain (i.e., pain most days over the past three months); (3) pain severity ≥ 3/10 on the Brief Pain Inventory (BPI) 79 , (4) reside in the United States; (4) English language proficiency; and (5) willing to participate in a follow-up survey. To recruit and perform an initial eligibility screen, MTurk workers responded to a single question regarding chronic pain status, which was embedded into other MTurk studies: “Have you had pain most days (more than half) over the past 3 months?” Those who indicated the presence of chronic pain were invited to complete additional questions to further evaluate eligibility. Specifically, they responded to questions regarding their country of residence and whether they were willing to be re-contacted to complete an additional survey in the future, as well as completed select questions from the Brief Pain Inventory (BPI 19 ). BPI questions were used to assess the presence of pain and to determine pain severity (average of current pain and worst, least and average pain in the past 24 hours). Because of a potential delay in the initial and subsequent screener, they were asked again whether they had experienced pain most days over the past three months. In addition to eligibility criteria, additional criteria were applied to maximize the overall data quality and reliability. Specifically, MTurk workers have “approval ratings” based on their responses to prior MTurk surveys or projects. In the current study, only MTurk workers with a minimal approval rating of 95% were invited to participate. We adopted additional measures to ensure quality data. Three attention check questions were embedded into the survey battery (e.g., “Please select Never True”). Data from participants who failed one or more attention checks were considered low quality and excluded from analysis. We established an a priori survey duration based on pre-launch pilot testing and expert consultation from Cloud Research. The T1 survey had an average completion duration of 35 minutes; data were excluded from participants who took substantially greater than (> 60 minutes) or less than (< 16 minutes) the average. The data from participants who did not meet our quality standards were excluded from analysis. Only participants who met quality standards at T1 were invited to complete the T2 survey battery. With regards to initial recruitment, a total of 30,096 MTurk workers completed the initial screening question, and of these people, 10,308 (34.3%) indicated the presence of chronic pain. These participants were invited to complete a second screener, of whom 2,153 (20.9%) completed the screener and met eligibility criteria. These participants were invited to participate in the study. Of those invited to participate, 1,809 (84%) initiated the study. Of these participants, 325 (18%) of participants provided data that did not meet data quality standards. Of the remaining 1484 who provided data that met data quality standards, we identified 31 cases in which a participant had completed the survey more than once. In this case, we deleted the second entry (unless the first entry was incomplete). This yielded a final T1 sample size of N = 1453. Regarding the T2 survey, participants in the final T1 sample (N = 1453) were invited to participate, and 884 (61% retention) initiated the survey. A total of six participants did not provide data (i.e., they initiated the survey but did not provide survey responses), yielding a final Time 2 sample size of N = 878. Participants responded to questions about demographic background, including age, gender, race, ethnicity, education, employment, income and marital status. Participants indicated the duration of their chronic pain and were presented with a list of chronic pain conditions and asked to indicate which they were currently experiencing. Participants completed the pain severity and pain interference subscales of The Brief Pain Inventory – Short Form (BPI 19 ). To assess pain severity, participants rated their current pain and worst, least, and average pain in the past 24 hours, using a 11-point Likert scale ranging from 0 ( no pain ) to 10 ( pain as bad as you can imagine ). To assess pain interference, participants indicated the extent to which pain interfered with seven domains of function, including general activity, mood, walking ability, normal work, relations with others, sleep, and enjoyment of life. These items are rated on an 11-point Likert scale ranging from 0 ( does not interfere ) to 10 ( complete interferes ). A mean score is calculated for pain severity and pain interference, and higher scores indicate more severe pain or greater interference. The BPI has demonstrated good internal consistency, test-retest reliability, criterion-related, and construct validity, and it is responsive to detecting change in pain over time in adults with chronic pain. 5 , 43 , 55 , 79 . In this study, internal consistency for pain severity was .71 at baseline and .90 at follow-up, and for pain interference, it was .91 at baseline and .92 at follow-up. Participants completed the Widespread Pain Index (WPI), a self-report measure used to assess the bodily distribution of pain. 18 , 87 Participants are presented with a body map annotated with 19 distinct bodily regions (e.g., right lower leg, left upper arm) and asked to indicate where they have experienced pain in the past seven days. Each indication is equal to a score of 1. Responses are summed, yielding a total score, with higher scores indicating greater widespread pain. The WPI was used in the current study to characterize the average number of pain locations in the sample. The Pain Catastrophizing Scale (PCS 75 ) is a 13-item self-report questionnaire that assesses pain-related cognitions in three domains: rumination, magnification, and helplessness. Participants complete four items assessing rumination (e.g., “I keep thinking about how much it hurts”), three items assessing magnification (e.g., “I become afraid that the pain will get worse”), and six items assessing helplessness (e.g., “There’s nothing I can do to reduce the intensity of the pain”). Items are rated on a 5-point Likert scale ranging from 0 ( not at all ) to 4 ( all the time ). A total score is yielded by summing the items, with scores ranging from 0 to 52, with higher scores indicating greater catastrophizing. Three subscales scores for rumination, magnification, and helplessness can also be formed, although in the current study we examined the total score only. The PCS has demonstrated good internal consistency, test-retest reliability, concurrent validity, and discriminant validity in adults with chronic pain. 64 , 65 , 86 Rumination, Magnification, and Helplessness subscales had Pearson rs ranging from .72 ~ .81. Internal consistency for the total PCS was .95. The 8-item version of the Chronic Pain Acceptance Questionnaire (CPAQ-8 26 ) is a self-report measure that assesses two pain acceptance domains, including activity engagement and pain willingness. Using a 7-point Likert scale ranging from 0 ( never ) to 6 ( always true ), participants rate the extent to which they agree with statements about pain acceptance, such as “I am getting on with the business of living no matter what my level of pain is” (activity engagement) and “Keeping my pain level under control takes first priority whenever I am doing something” (pain willingness). Subscale scores can be generated from this measure; however, in the currently study, we only examined the total score. Total scores range from 0 to 48, with higher scores represent greater acceptance. The CPAQ-8 has demonstrated adequate internal reliability, test-retest reliability, and construct validity in adults with chronic pain. 25 , 26 Activity Engagement and Pain Willingness subscales had Pearson r of .32. Internal consistency was .82 for the total score. The Toronto Alexithymia Scale 20-Item (TAS-20 7 ) is a self-report questionnaire that assesses alexithymia. The TAS-20 consists of three subscales: difficulty identifying feelings (DIF; 7 items), difficulty describing feelings (DDF; 5 items), and externally oriented thinking (EOT; 8 items). Participants report the degree to which they agreed with alexithymia-related statements, such as “I am often confused about what emotion I am feeling,” “I am able to describe my feelings easily,” and “I prefer to just let things happen rather than to understand why they turned out that way.” Items are rated on a 5-point Likert scale ranging from 1 ( strongly disagree ) to 5 ( strongly agree ). A total score and subscale scores are calculated by summing the items, with higher scores indicating higher levels of alexithymia. The TAS-20 is the most commonly used measure for assessing alexithymia in individuals with chronic pain. 2 TAS-20 total scores have demonstrated adequate internal consistency and test-retest reliability in adults with pain. 15 , 71 However, the psychometric properties of the EOT subscale of the TAS-20 specifically are weak, 46 particularly among individuals from non-Western countries. 20 Some have argued that EOT represents a fundamentally different construct, which should not be assessed in aggregate with DIF and DDF subscales. 46 Results from meta-analyses suggest that unlike DIF and DDF, EOT is not significantly higher in individuals with chronic pain compared to those without. 2 In the current study, internal consistency for EOT was poor (α = .60) whereas internal consistency for DIF (α = .88) and DDF (α = .81) was good. For these reasons, we excluded the EOT subscale from the present analyses. We instead examined an aggregate score comprised of DIF and DDF only (hereafter referred to as “TAS-12” reflecting the total DIF and DDF items). DIF and DDF subscales had Pearson r of .81. The Emotion Regulation Questionnaire (ERQ) is a 10-item self-report measure that assesses the use of emotion regulation strategies. 35 The questionnaire consists of two subscales: cognitive reappraisal and expressive suppression. Participants respond to four items assessing cognitive reappraisal (e.g., When I want to feel more positive emotion (such as joy or amusement), I change what I’m thinking about ,”) and six items assessing expressive suppression (e.g., “I control my emotions by not expressing them .”). Items are rated on a 7-point Likert scale ranging from 1 ( strongly disagree ) to 7 ( strongly agree ). Subscale scores are calculated by averaging the scores for all subscale items, with higher scores representing greater use of reappraisal or suppression. These subscales are considered independent and not intended to be examined as an aggregate. For this reason, we examine ERQ subscale scores individually in the current study. The ERQ has demonstrated good internal consistency in adults with pain. 28 , 82 , 88 In this study, internal consistency was .90 for cognitive reappraisal and .83 for expressive suppression. The Emotional Approach Coping Scale (EAC-8 73 ) is an 8-item measure that assesses a respondent’s approach to coping with stress. There are two subscales: emotional processing (i.e., the acknowledgment, understanding, and acceptance of one’s emotions) and emotional expression (i.e., the verbal or nonverbal disclosure of one’s emotions). Emotion processing items include, for example, “I take time to figure out what I am feeling.” Emotional expression items include, for example: “I take time to express my emotions.” Participants responded using a 4-point Likert scale ranging from 1 ( I don’t do this at all ) to 4 ( I do this a lot ). In the current study, the total score was used. Total scores range from 8 to 32, with higher scores indicating more adaptive emotional coping. The EAC total and subscales have demonstrated good to excellent internal consistency in adults with chronic pain. 71 , 72 , 91 Emotion Process and Emotion Expression subscales had Pearson r or .62. In this study, internal consistency was .90 for the total score. First, descriptive statistics were computed to summarize the baseline characteristics of the current sample. Second, attrition analyses were conducted on key socio-demographic and pain-related variables at baseline. Third, multicollinearity was examined among study variables based upon bivariate correlations, tolerance (cutoff = .10 or less) and VIF (cutoff = 10 or more) 78 . Fourth, using the Mplus software (Version 8.2 60 ), we conducted two path models that examine predictors of pain severity and interference that are assessed at three-month follow-up simultaneously: (1) Model 1 includes only pain-specific emotion regulation predictors; and (2) Model 2 (final model) includes both pain-specific and general emotion regulation predictors. Model 2 tests the predictive effects of general emotion regulation on pain severity and interference at three-month follow-up over and above the effects of pain-specific emotion regulation. For each model, independent variables were total sum questionnaire scores, unless otherwise indicated based on psychometric properties. Several baseline socio-demographic variables (i.e., age, gender, race, ethnicity, income, and education) that have previously been reported to be associated with pain-related outcomes, 16 , 38 , 58 and baseline pain severity and interference were included as covariates in both models. For the race dummy covariates, we set Black/African American as the reference group, as numerous studies revealed pain disparities among individuals who identify as Black/African American. 16 , 38 For gender, we focused on the binary gender category (i.e., male and female) in both models, as a very small proportion (0.5%) of individuals identified themselves as non-binary/genderqueer. Finally, to further scrutinize our models in light of TAS-20 limitations described in the Questionnaires section, we conducted sensitivity analyses to examine the effect of Model 1 and Model 2 when (1) the full TAS-20 (versus TAS-12) was included as a predictor, and (2) isolating specific TAS-20 subscales to parse out sensory focused items. To determine model fit of proposed path models, we used the Comparative Fit Index (CFI; critical value ≥ .90; [1]), Standardized Root Mean Residual (SRMR; critical value, ≤ .10; 44 ), and the Root Mean Squared Estimate of Approximation (RMSEA; critical value ≤ .08; 74 ). Missing data were handled by the full information maximum likelihood (FIML) method, which is superior to other traditional missing data handling methods (e.g., listwise deletion, mean imputation, etc.), based on the missing at random (MAR) assumption (i.e., the missingness does not depend on the value of outcome itself, but related to some of the observed data). 24 In terms of effect sizes, we reported R 2 values for each endogenous (outcome) variable in path models and provided standardized regression path estimates. We computed the sensitivity to detect regression effects based upon the current sample size. Power analysis conducted using G*Power software revealed that in a multiple regression model with 15 predictors, a sample of 1453 with alpha level of .05 (two-tailed) can produce .80 statistical power to detect even very small effects (Cohen’s d = .01).