Secondary Logo

Journal Logo

Pain-related gaze biases and later functioning among adults with chronic pain: a longitudinal eye-tracking study

Jackson, Todda,b,*; Yang, Zhoua; Su, Lina

doi: 10.1097/j.pain.0000000000001614
Research Paper
Editor's Choice
Global Year 2019

In previous studies that examined the impact of attention biases (ABs) on later pain outcomes, reaction times (RTs) in response to brief stimulus presentations had been used as measures of attention. Consequently, little is known about effects of ABs assessed during presentations of cues or biases in prolonged attention towards pain stimuli as influences on subsequent functioning. To address these gaps, 89 adults with chronic pain (68 women, 21 men) engaged in a baseline dot-probe task in which visual attention was tracked during injury-neutral (I-N) image pair presentations as well as a 6-month follow-up reassessing pain intensity and interference from pain. Neither RTs to probes after image pair offsets nor biases in initial orienting of gaze towards injury images predicted follow-up outcomes. However, participants who gazed at injury images for longer durations during I-N trials reported significantly more pain and interference at follow-up than did peers who gazed at injury images for less time, even after the impact of other significant baseline predictors had been controlled. In sum, results provided initial evidence for gaze biases reflecting prolonged vigilance towards pain-related information as a potential risk factor for relative elevations in pain and interference from chronic pain.

In this longitudinal eye-tracking study, prolonged gaze toward pain images during baseline dot-probe trials independently predicted elevations in pain and interference at 6-month follow-up.

aKey Laboratory of Cognition and Personality, Southwest University, Chongqing, China

bDepartment of Psychology, University of Macau, Taipa, Macau S.A.R.

*Corresponding author. Address: Key Laboratory of Cognition and Personality, Southwest University, Chongqing 400715, China. Tel.: 86-13883224482. E-mail address: (T. Jackson).

Sponsorships or competing interests that may be relevant to content are disclosed at the end of this article.

Back to Top | Article Outline

1. Introduction

Attention biases (ABs) during exposure to pain-related information have been implicated as key influences on the development and maintenance of chronic pain.3,25,26 For example, Eccleston and Crombez3 argue that people are predisposed to prioritize pain as a focus of attention because it facilitates escape, yet the capacity to shift attention towards other tasks or demands can mitigate negative effects of pain. Unfortunately, when pain becomes chronic, competing environmental demands are potentially reduced, and pain is more likely to capture and sustain attention, hence perpetuating discomfort, disability, and other consequences.

Initial longitudinal studies found variable relations between ABs from laboratory tasks before surgery and postoperative pain intensity.8–10,15 In subsequent research, Sharpe et al.20 found baseline ABs away from affective pain words predicted chronicity of subacute or acute back pain at 3- and 6-month follow-ups. Furthermore, baseline ABs related to sensory and affective pain words had significant bivariate relations with disability levels at the 6-month follow-up, albeit effects of these ABs were not significant in a multivariate prediction model that included other baseline influences.

In the sole prospective study on chronic pain,28 pain-related ABs displayed during a baseline spatial cuing task were not related to daily pain severity or disability ratings from a 2-week follow-up after controlling for other significant baseline factors. However, in the prediction of disability, ABs had a unique moderating impact wherein the association between daily pain severity and daily disability was stronger among patients who displayed stronger baseline ABs toward pain-related information.

Although these studies underscore the potential utility of assessing relations between ABs shown during laboratory tasks and subsequent pain outcomes, this literature has important gaps. Most critically, all past studies assessed ABs based on reaction times (RTs) during brief stimulus presentation durations (100-500 ms). As such, initial orienting of attention was examined to the neglect of attention maintenance, despite emphasis on ABs reflecting prolonged vigilance or disengagement difficulties in related theory.3,26 Reviewers have also expressed concerns that RTs provide static, imprecise AB indexes compared with alternative strategies that permit AB assessments during pain cue presentations.1,18 Because eye movements can be guided by attention, tracking of gaze during pain-related vs nonpainful image presentations allows for direct, continuous measurement of visual-spatial attention that addresses limitations of ABs inferred from RTs.

While longitudinal eye-tracking studies have not been undertaken in chronic pain samples, threatening pain cues tend to elicit more overall gaze time during exposure than do less threatening alternatives.4,6,7 Arguably, then, ABs reflecting vigilance in orienting and/or maintenance of gaze towards pain-related cues that are relatively benign (ie, visual depictions) may predict poorer outcomes of chronic pain. To evaluate this premise, we assessed effects of biases in orienting and maintenance of gaze towards pain images from a dot-probe task on pain intensity and interference levels of adults with chronic pain at a 6-month follow-up. Following associated RT studies,20,28 unique effects of significant gaze indices were examined after controlling for other significant baseline correlates of follow-up outcomes.

Back to Top | Article Outline

2. Method

2.1. Participants

The final sample (68 women, 21 men) comprised adults with ongoing pain every day or most days for the past 3 months or longer. Participants ranged in age from 18 to 55 years (M = 26.70 years, SD = 10.07 years) with a majority reporting at least some post–high school education (85%). The average pain duration was 37.67 months (SD = 58.11 months, range: 3-462 months); however, pain durations of 4 participants were more than 3 SDs above the mean and were replaced with the sample average in main analyses. Primary pain sites included neck or shoulder (46%), low back (28%), extremity (13%), head or face (10%), and other (3%). A majority of participants (64%) reported more than one pain site (M = 1.89, SD = 1.11). Most reported pain every day (84%) although a minority (42%) acknowledged current analgesic use for pain. All participants had normal or corrected-to-normal vision and reported no current/past neurological disease or major psychiatric illness (ie, schizophrenia, psychotic disorder, bipolar disorder, or dementia) that might interfere with comprehension.

Back to Top | Article Outline

2.2. Procedure

The research was approved by the Human Research Ethics Committee of Southwest University, Chongqing. Participants were recruited for an ongoing study on chronic pain in China through contacts and printed advertisements at the university hospital, a community-based hospital, university-affiliated housing facilities, and the campus electronic bulletin board. Selection criteria included current noncancer pain experienced most days or every day for 3 months or longer, normal or corrected-to-normal vision, right-handedness, literacy with printed Chinese language, absence of prescription medication use (eg, opioids), and absence of neurological or psychiatric illness that interfered with comprehension. Before the baseline assessment, 454 people with chronic pain who met these selection criteria had completed an initial questionnaire study. Upon completing that study, respondents were asked whether they would like to participate in other studies from our laboratory. Of those who expressed interest (N = 105), 98 volunteers were available and completed the current study 2 to 6 weeks later.

After arriving at their scheduled appointments, participants signed an informed consent that introduced the main research focus (“to examine how reactions to pictures from a laboratory task are related to other life experiences”), tasks (ie, viewing painful and nonpainful images while having eye movements tracked, completing several questionnaires on current experiences, and a brief telephone follow-up), the voluntary nature of consent, time involved (about 40 minutes), and compensation (100 RMB). Subsequently, they completed a dot-probe task and measures related to current pain experiences. Data from 5 participants (3 men) were lost due to problems during eye tracking. Six months later, participants completed a telephone interview featuring a reassessment of pain intensity and pain interference items used in the baseline assessment and were debriefed about the main research focus. The follow-up resulted in additional missing data from 4 men who had lost contact with the study.

Back to Top | Article Outline

2.3. Image set

For the dot-probe task, we used 64 digital color images from other published research.13,14 Images depicting a model's hand, forearm, or foot in injury situations (16 images) or similar neutral situations (48 images) were used to create 16 injury-neutral (I-N) image pairs and 16 neutral-contrast-neutral (Nc-N) image pairs. Following related eye-tracking studies,6,7 Nc-N pairs were created by selecting 16 neutral images at random and designating them as Nc images, each to be paired with 1 of 16 remaining neutral (N) images. Nc-N image pairs served primarily as fillers but were also used for additional baseline comparisons. All injury images featured depictions of possible or actual tissue damage that can occur in everyday life (eg, finger cut while slicing vegetables and hand caught in closed door) and typically result in pain. Neutral pictures corresponded to those shown in injury depictions without a nociceptive component (eg, hand slicing vegetables and hand resting on top of closed door). Each picture had dimensions of 11 × 10 cm (width × height) and 100 pixels per inch. The distance between the 2 images forming a pair was 10 cm. Images were vertically centered against a black background.

Each image pair was matched for luminance, contrast, color, and perceived motion. Injury images evoked significantly higher pain intensity, negative valence, and arousal ratings in validation research.13 In this study, participants rated how much fear the injury and neutral images evoked from “0 = no fear at all” to “9 = extreme fear” following the dot-probe task. As expected, injury images elicited more fear (M = 5.41, SD = 1.77) than did neutral images (M = 1.49, SD = 0.58), P < 0.001, = 0.86.

Back to Top | Article Outline

2.4. Dot-probe task

Upon commencing the dot-probe task, participants were informed onscreen as well as verbally that the upcoming task involved attending to I-N and Nc-N image pairs then identifying the location of the dot probe that appeared in the place of 1 image as quickly and accurately as possible. During image pair presentations, eye movements were recorded. Trials began with a central fixation cross “+” shown for 1000 ms, followed by an image pair that appeared for 2000 ms, in line with past eye-movement research.1,5,6,16,19,31,32 Immediately after image pair offsets, a probe () appeared in the location formerly occupied by one of the images. Participants had to indicate the location by pressing the A or L keys of a standard keyboard to denote left vs right locations, respectively. Each probe appeared until a response was made or for a 4-second maximum. Probes were equated to appear in congruent and incongruent positions after I-N and Nc-N image pairs, respectively.

Before the formal task, 8 practice trials (repeatable if desired) were undertaken with different images to increase familiarity with trial sequence requirements. The formal dot-probe task featured 2 I-N blocks and 2 Nc-N blocks of 32 trials each. Within each I-N block, 16 I-N image pairs were randomly presented twice, with each image appearing once on each side of the screen. Sixteen probes following image pairs were presented in positions that were congruent with injury images (8 on right and 8 on left) while 16 probes appeared in positions incongruent with these images (8 on right and 8 on left). Nc-N blocks comprised 16 neutral (Nc-N) image pairs presented following guidelines used for I-N block. Presentation order of I-N and Nc-N blocks was counterbalanced between and within participants. A 1 minute break followed each block.

Back to Top | Article Outline

2.5. Self-report measures

2.5.1. Chronic pain grade

Pain intensity levels at baseline and 6-month follow-up were calculated by summing responses on chronic pain grade (CPG) items assessing (1) “pain right now” as well as (2) worst pain and (3) average pain during the past 3 months.22 Each pain intensity item was rated from 0 (no pain) to 10 (pain as bad as could be). Pain interference levels at baseline and follow-up were calculated from sums of CPG items assessing impairment in daily activities, recreational activities, social and family activities, and ability to study/work, each of which was rated from 0 (no interference) to 10 (unable to carry on any activities). Chronic pain grade dimensions have satisfactory psychometrics in samples from China.30,34 In this sample, alphas for pain intensity (baseline: α = 0.89, 6-month follow-up: α = 0.82) and pain interference (baseline: α = 0.79, 6-month follow-up α = 0.76) were acceptable.

Back to Top | Article Outline

2.5.2. Fear of Pain Questionnaire–Chinese

The 25-item Fear of Pain Questionnaire–Chinese (FPQ-C) includes subscales that assess fear of (1) minor pain, (2) severe pain, and (3) medical pain. Items were rated from “1 = no fear at all” to “5 = extreme fear.”31,32 The FPQ-C was derived from a factor analysis of Fear of Pain Questionnaire–III items12 and has acceptable reliability and construct validity in Chinese samples.31,32 In this study, the FPQ-C alpha was α = 0.95.

Back to Top | Article Outline

2.5.3. Pain catastrophizing scale

The 13-item pain catastrophizing scale (PCS) assesses features of pain catastrophizing including rumination, magnification, and helplessness.23 Items were rated from 0 (never like that) to 4 (always like that) and summed to calculate total scores. The PCS factor structure and validity have support in Chinese samples.33 In this study, the total PCS alpha was α = 0.94.

Back to Top | Article Outline

2.5.4. Center for epidemiologic studies depression

The 20-item Center for Epidemiologic Studies Depression (CES-D) assessed the frequency of specific depressive symptoms experienced during the past week on a scale anchored by 0 = rarely or none of the time (less than 1 day) and 3 = most of all the times (5-7 days).17 CES-D factor structure, reliability, and validity have been supported in Chinese samples.34 In this sample, the CES-D alpha was α = 0.93.

Back to Top | Article Outline

2.5.5. Demographics

Age, sex, highest education level (primary school completion or lower, middle school completion/partial completion, high school completion/partial completion or postsecondary education), primary pain site, pain duration, pain every day (no vs yes), and current analgesic medication use (no vs yes) were assessed.

Back to Top | Article Outline

2.6. Eye-tracking apparatus

Image pairs were presented on a 19-inch, 85-Hz monitor connected to a Pentium IV 3.2-GHz computer. The EyeLink 1000 tracker (SR Research, Mississauga, Canada) was linked to a Pentium IV 2.8-GHz host computer and recorded eye movements. The eye-tracker sampling rate was 500 Hz with spatial accuracy greater than 0.5°, with 0.01° resolution in the pupil-tracking mode. An infrared motion system tracked eye movements. A forehead and chin rest kept viewing distance constant and minimized head movements. Participants sat 70 cm from the monitor screen resulting in a 29° horizontal × 22° vertical visual field. Before the task, a standardized calibration procedure for the assessment of eye movements was conducted by requesting participants focus on 9 white dots randomly appearing on the black display space in a 3 × 3 array (from the top left to the bottom right of the screen). EM data were recorded during each trial, starting just before image pair onsets and terminating immediately after their offsets.

Back to Top | Article Outline

2.7. Data preparation

2.7.1. Gaze bias indices

Saccades that remained stable within a 1° visual angle for at least 100 ms were defined as fixations, the position and duration of which were recorded. Fixations on either image in each pair were identified when (1) participants fixated in the central region before image-pair onsets, (2) saccades occurred at least 100 ms after pair onsets and before pair offsets (fixations with shorter latencies were unlikely to be contingent on images and may have reflected express saccades or anticipatory eye movements instead), and (3) participants fixated on 1 image, not the space between images, during image presentations. In this sample, first fixations were made on one image in 96% of trials.

First fixation direction biases for injury images and neutral images were obtained by calculating the percentage of trials in which gaze initially fixated on these images rather than complementary neutral images. Overall, gaze duration biases were based on total gaze time from each trial (2000 ms) that participants fixated upon injury images in I-N image pairs and neutral images in Nc-N image pairs.

Back to Top | Article Outline

2.7.2. Reaction time bias scores

Reaction times were based on the interval between image pair offsets and probe location responses. In calculating RT bias scores, trials with incorrect probe location responses (<0.02%) or postoffset RTs more than 3 SDs above the mean (>1602 ms) or < 200 ms were deleted as an initial step. Reaction time bias scores to injury images were calculated by subtracting RT on incongruent trials (probe − N image) from RT on congruent trials (probe − I image). Reaction time bias scores for Nc images were calculated by subtracting RT on congruent trials (probe − N image) from RT on incongruent trials (probe − Nc) of neutral image pairs.

Back to Top | Article Outline

2.8. Design and data analysis

Analyses were performed with SPSS 20.0. Because volunteers were enlisted from different settings, preliminary analyses were run to test recruitment channel (university vs community) differences on research measures through chi-square tests, univariate analyses of variance (ANOVAs), and repeated-measures ANOVAs. Next, effects of image type on orienting and maintenance of gaze during dot-probe task trials were examined. First fixation direction biases for I-N and Nc-N image pairs were examined through one-sample t-tests compared with a 50% chance level, following other published work.6,7,16 Paired-sample t-tests assessed overall gaze durations toward each image type in I-N and Nc-N image pair blocks using operational definitions below. The effect size measure for gaze response differences was Cohen's d based on a formula that corrects for dependence between means for within-subject analyses.6,7 Reaction times for each trial were submitted to a 2 (image pair type: I-Np vs Nc-N) × 2 (probe location: congruent vs incongruent) repeated-measures ANOVA. Image pair type and probe location were within-subject factors. Partial eta-squared () was the effect size measure.

Regarding the main research focus, initial bivariate correlation analyses evaluated strengths of relation between responses on baseline measures (demographics, psychological measures, AB indices, pain intensity, and interference) and follow-up levels of pain intensity and interference. Subsequently, significant non-multicollinear baseline correlates of follow-up outcomes (r > 0.70) were assessed within separate hierarchical standard multiple regression models for follow-up (1) pain intensity and (2) interference. In each model, baseline pain intensity (or interference) and significant demographics (block 1), significant psychological influences (block 2), and significant AB indices (block 3) were possible predictors. As a result, the unique impact of salient ABs from the baseline dot-probe task on follow-up outcomes could be evaluated, independent of all other identified predictors. In these analyses, effects were interpreted as significant using the conventional P < 0.05 value (2-tailed). A minimum sample N of 82 was estimated based on recommendations that multivariate regression models include a minimum N of 50 + 8 subjects for each predictor,24 and (2) the assumption that each model would include 4 to 5 predictors.

Back to Top | Article Outline

3. Results

3.1. Preliminary analyses

3.1.1. Recruitment channel differences on research measures

Chi-square tests revealed no recruitment channel differences on categorical measures of sex, primary pain site, current analgesic use, or pain experienced every day (all P's > 0.38). Analyses of variances indicated there was a significant recruitment channel difference on age [university: M = 22.38 years, SD = 2.21 vs community: M = 32.76 years, SD = 13.29, F (1, 87) = 30.64, P < 0.001, η2 = 0.26] and a statistical trend for pain duration [university: M = 27.65 months, SD = 30.61 vs community: M = 51.76 months, SD = 81.10, F (1, 87) = 3.84, P < 0.053, η2 = 0.04]. No recruitment channel differences were found on total pain sites, baseline measures of depression, fear of pain, pain catastrophizing, or any gaze bias index (all P's >0.17).

In a repeated-measures ANOVA for pain intensity, effects of recruitment channel, F (1, 87) = 0.00, P > 0.96, and its interaction with study phase, F (1, 87) = 0.97, P > 0.32, were not significant, although the overall sample reported a significant reduction in pain intensity from baseline (M = 12.66, SE = 0.52) to follow-up (M = 10.37, SE = 0.50), F (1, 87) = 17.56, P < 0.001, = 0.17. For pain interference, there was a recruitment channel difference reflecting significantly less overall interference within the university subgroup (M = 13.72, SE = 0.49) than the community subgroup (M = 15.51, SE = 0.58), F (1, 87) = 5.52, P < 0.021, = 0.06. In addition, the overall sample reported a significant decrease in pain interference from baseline (M = 15.68, SE = 0.45) to follow-up (M = 13.56, SE = 0.46), F (1, 87) = 17.60, P < 0.001, = 0.17, although the recruitment channel × study phase interaction was not significant, F (1, 87) = 2.21, P > 0.14.

Back to Top | Article Outline

3.1.2. Image-type differences in orienting and maintenance of gaze

With respect to orienting of gaze (ie, first fixation direction biases), the sample more often fixated initially upon injury images in I-N pairs (M = 0.56, SD = 0.06), t (88) = 8.34, P < 0.001, d = 0.94. Conversely, first fixations on arbitrarily selected neutral images from Nc-N image pairs did not exceed a chance level (M = 0.50, SD = 0.06), F (1, 88) = −0.44, P < 0.661, d = −0.05.

On the attention maintenance index, participants displayed significantly longer overall gaze durations toward injury images from I-N image pairs (P images: M = 923.09 ms, SD = 114.15 vs N images: M = 723.24 ms, SD = 107.79), t (88) = 9.48, P < 0.001, d = 0.98. By contrast, overall gaze biases were not evident for Nc-N image pairs (Nc images: M = 823.79 ms, SD = 56.47 vs N images: M = 814.32 ms, SD = 52.69), t (88) = 1.38, P = 0.172, d = 0.14.

Back to Top | Article Outline

3.1.3. Effects of image type and probe location on reaction times

For RTs in identifying probe locations, nonsignificant effects were found for image pair type [I-N image pairs: M = 611.51 ms, SD = 128.83 vs Nc-N image pairs: M = 607.44 ms, SD = 126.16, F (1, 88) = 0.67, P = 0.42, = 0.01], probe location [congruent: M = 611.41 ms, SD = 128.3 vs incongruent: M = 607.53 ms, SD = 126.67, F (1, 88) = 0.76, P = 0.39, = 0.01], and their interaction, F (1, 88) = 0.78, P = 0.38, = 0.01.

Back to Top | Article Outline

3.2. Main analyses

3.2.1. Bivariate relations of baseline measures with follow-up pain outcomes

Bivariate relations of baseline demographics (age, sex, education, pain duration, total pain sites, and analgesic use) and baseline gaze or RT bias scores from Nc-N image pair trials with follow-up pain intensity scores were not significant (P's > 0.17). Table 1 summarizes bivariate correlations of baseline responses on psychological measures and pain-related AB indices with follow-up pain intensity. As shown there, follow-up pain intensity scores had significant positive relations with baseline levels of pain intensity, interference, and pain catastrophizing as well as overall gaze biases towards injury images during the dot-probe task. Follow-up pain intensity was not related to fear of pain, depression, or ABs based on orienting of gaze or RTs during I-N pair presentations (Table 1).

Table 1

Table 1

Follow-up interference scores were not related to participant sex, pain duration, total pain sites, or analgesic use from baseline (all P's > 0.28). However, older participants (r = 0.33, P = 0.002) and those with less education (r = −0.25, P = 0.019) reported more interference from pain at follow-up. Correlations of gaze and RT bias scores from Nc-N pair trials with follow-up interference were not significant (all P's > 0.23). However, baseline interference and pain intensity scores, pain catastrophizing scores, and overall gaze duration biases toward injury images in I-N trials had significant associations with follow-up interference (Table 1).

Back to Top | Article Outline

3.2.2. Multivariate prediction models for follow-up pain outcomes

In an initial standard hierarchical multiple regression model, baseline pain intensity scores (block 1), pain catastrophizing scores (block 2), and gaze biases toward injury images (block 3) were included as predictors while baseline interference was excluded because of multicollinearity with pain intensity. The overall model explained adjusted R2 = 0.216 of the variance in follow-up pain intensity, F (3, 85) = 9.08, P < 0.001. Participants who reported more intense pain at baseline tended to report more pain 6 months later too. The impact of pain catastrophizing was not significant in the multivariate model (Table 2). Critically, after controlling for these measures, longer overall gaze durations at injury images during baseline dot probe trials accounted for significant additional variance in follow-up pain intensity (Table 2).

Table 2

Table 2

In the prediction model for follow-up interference from pain, baseline interference scores and age were included in block 1 while pain intensity and education were excluded because of multicollinear relations with these predictors. Pain catastrophizing and overall gaze duration biases toward injury images were entered in blocks 2 and 3, respectively. Together, these measures explained adjusted R2 = 0.280 of the total variance in follow-up interference levels, F (4, 84) = 9.57, P < 0.001. Baseline elevations in interference from pain and older age made unique contributions to follow-up interference levels while the impact of pain catastrophizing was not significant (Table 2). After controlling for these measures, stronger overall gaze duration biases toward injury images from the baseline dot-probe task contributed to the prediction of elevations in interference at follow-up (Table 2).

Back to Top | Article Outline

3.2.3. Supplementary analyses of gaze duration—pain outcome associations

Supplementary multiple regression analyses evaluated the extent to which gaze duration bias score results varied as a function of analysis approach and measures included in prediction models. In an alternate standard hierarchical regression model for follow-up pain intensity that included all significant baseline correlates, the impact of overall gaze duration biases remained significant (β = 0.28, t = 2.91, P = 0.005, R2 Change = 0.076), after controlling for baseline pain intensity, interference, and pain catastrophizing scores. On follow-up pain interference, overall gaze duration bias scores had a unique impact (β = 0.25, t = 2.61, P = 0.011, R2 Change = 0.061), after controlling for all other significant baseline correlates (age, education and baseline pain intensity, interference, and pain catastrophizing scores).

Alternately, we ran stepwise multiple regression analyses that retained only significant unique predictors within final models. With this approach, the only significant predictors of follow-up pain intensity were baseline pain intensity and gaze duration bias scores (β = 0.28, t = 2.86, P < 0.01, R2 Change = 0.075). For follow-up interference, significant predictors were baseline interference, age, and gaze duration bias scores (β = 0.22, t = 2.33, P < 0.05, R2 change = 0.046). In sum, similar results from alternative analysis strategies indicated gaze duration bias results were not due to analysis approach or measures included in prediction models.

Because of recruitment channel differences (age and pain duration) and reviewer comments, supplementary analyses were also run to moderating effects of age and pain duration on gaze duration bias—follow-up pain outcome relations. Follow-up pain intensity had a near-significant correlation with age × gaze duration bias scores (P = 0.051). In the related multivariate prediction model, the impact of centered age × gaze duration bias scores was attenuated further (β = 0.175, t = 1.79, P = 0.077), after controlling for baseline pain intensity, catastrophizing, and gaze duration bias scores; nonetheless, this trend reflected stronger relations of gaze duration bias scores with follow-up pain intensity in the younger (β = 0.34, t = 2.70, P = 0.011) compared with the older (β = 0.20, t = 1.24, P = 0.224) than median age subgroup. Follow-up pain intensity scores were not related to pain duration × gaze duration bias scores (P = 0.293), and follow-up interference was not related to age × gaze duration bias (P = 0.455) or pain duration × gaze duration bias (P = 0.090) interactions.

Back to Top | Article Outline

4. Discussion

To the best of our knowledge, this study is the first to evaluate individual differences in image-related gaze biases as predictors of later functioning among adults with chronic pain. Participants who displayed longer overall dwell times upon injury images during an initial dot-probe task reported more pain and interference from pain in a 6-month follow-up than did cohorts who had shorter overall dwell times upon injury images. The impact of distal gaze duration biases was significant after first controlling for baseline levels of pain intensity or interference and other significant baseline correlates of follow-up outcomes. By contrast, neither baseline biases in orienting of gaze towards injury images nor RTs to probes that followed image pair offsets predicted follow-up outcomes. Implications of main findings are discussed below.

Cognitive-behavioral theorists have posited that vigilance for pain-related information and/or difficulties shifting attention towards other tasks can perpetuate discomfort and disability from chronic pain.3,26,29 Consistent with these premises, participants who displayed prolonged vigilance towards pain-related images from dot-probe trails subsequently experienced more pain and interference from pain than did peers who showed less vigilance. While vigilance can increase when the threat intensity of pain cues is high,4,6,7 pain-related images from this study were comparatively benign, and in that, they reflected typically minor injuries that occur in daily life and did not signal impending painful stimulation or reflect actual physical discomfort. As such, it can be hypothesized that some of those who showed heightened vigilance towards visual cues for injury during dot-probe trials may have had similar difficulties disengaging from external and internal pain cues present in their daily activities; for them, comparatively larger allocations of attentional resources towards pain-related cues rather than nonpainful alternatives may have contributed to relatively poor follow-up outcomes. Conversely, the tendency to shift attention away from external pain cues towards nonpainful cues during the laboratory task may have reflected and/or fostered comparatively resilient responses to chronic pain.

Supplementary analyses indicated the impact of overall gaze biases was not a function of analysis strategy or respondent differences in pain duration. However, moderator analyses for age revealed a statistical trend indicating associations between baseline gaze biases and follow-up pain intensity levels were statistically significant among younger participants but not their older peers. Given this trend, future prospective work should also consider possible moderating effects of age and time since pain onset on both (1) subsequent functioning in samples with ongoing pain and (2) responsiveness to AB modification interventions in these groups.

The absence of significant associations between overall gaze duration bias scores and baseline levels of pain intensity and interference is consistent with nonsignificant relations between ABs based on RT and pain intensity reported in a meta-analysis of cross-sectional studies.2 However, significant associations between baseline biases in overall gaze and follow-up pain outcomes dovetail with the pattern reported by Sharpe et al.20; in that study, baseline RT biases for word stimuli were not related to baseline disability levels but had significant correlations with disability scores at a 6-month follow-up. In tandem, these studies suggest that, over time, individual differences in ABs toward external pain cues can have cumulative effects on individual differences in trajectories of pain intensity and interference from pain. Because of the paucity of other published work on prospective associations of ABs with changes in chronic pain outcomes over extended intervals, this contention is tentative. Nonetheless, associated results provide an impetus for future extensions.

In contrast to links between baseline vigilance towards injury images and subsequent functioning, there was no evidence that initial orienting of gaze towards or away from these images was related to follow-up pain intensity and interference levels. As such, ABs reflecting relatively automatic orienting of gaze may be less relevant to later functioning than are ABs that reflect maintenance of gaze. Furthermore, recent evidence of lower reliability in the measurement of biases in orienting than maintenance of gaze21 suggests that weaker psychometrics of orienting indices may have also contributed to these null effects.

The absence of significant effects from baseline RT biases is seemingly inconsistent with previous longitudinal AB studies based on response times. To some extent, variable results reflect differences in samples and stimuli, given that only one previous study evaluated persons who already had chronic pain at baseline,28 all cited prospective RT studies used word rather than image stimuli, and word-type effects varied between past studies. More critically, because the current RT results were based on long stimulus durations reflecting maintenance of attention, they are not directly comparable with past RT studies featuring stimulus durations (ie, 100-500 ms) that constrained AB assessments to early processing stages.

Pain-related fear has had a prominent role in theory on pain-related ABs,29 and select eye-tracking studies have reported links between FOP and ABs based on pain word stimuli.19,31,32 However, neither trait fear of pain nor reported intensity of fear elicited by injury images was related to ABs assessed in this study. These results align with the absence of significant overall associations found in a meta-analysis of RT studies2 and methodologically similar studies that also reported nonsignificant relations between fear of pain and gaze biases towards injury images in healthy respondents and those with chronic pain.6,7 Although image category differences (ie, word vs injury) and the tendency to assess fear responses before or after rather than during laboratory task trials may help to explain contradictory associations, examining the impact of situational manipulations of pain-related threat4,6,7 vs subjective fear experiences on pain-related ABs merits more consideration in future work.

Finally, notwithstanding possible implications of individual differences in baseline gaze duration biases, it is not clear why the overall sample reported significant decreases in pain and interference from baseline to 6-month follow-up because no intervention was included in the current research design. However, the mean age of respondents was substantially younger than the population from which they were drawn,5 and younger participants tended to have shorter pain durations in addition to significantly less reported interference from pain at the 6-month follow-up. As such, overall improvements in outcomes for the sample may have been partial reflections of age and/or current involvement of many younger respondents in meaningful pursuits related to future well-being (ie, higher education).

Despite its implications, several limitations of this study warrant mention. First, because findings are based on a relatively young chronic pain sample, caution is needed in making generalizations to older groups and those with acute pain or postsurgical pain, which is operationalized more narrowly (eg, “current pain” or “pain at this moment”) than it was in this study. Second, although fear elicited by each image was assessed, other factors such as personal relevance of images and current mood were not measured and can influence attention.11,27 Third, results underscored the prospective impact of pain-related gaze biases, but it is not clear whether biases specific to injury images depicting tissue damage are the only ones relevant to changes in pain intensity and interference. Meta-analyses have found RT biases toward sensory pain words distinguish chronic pain samples from controls.2,27 Extensions that include sensory pain word stimuli and other image categories (eg, varied facial expressions and injury depictions) as well as uncomfortable somatosensory stimulation can clarify the impact of ABs toward different types of external cues and painful stimulation itself on later outcomes of chronic pain. Fourth, following from Skinner et al.21 who recently found test–retest reliability support for gaze bias indices during a 90-minute experimental session, prospective extensions should examine how stable gaze biases are over longer periods, per research on RT biases.20 Finally, while this study was included in a successful external grant application, ideally research should also be formally preregistered before formal data collection.

To the best of our knowledge, this study is the first to evaluate the status of distal biases in both orienting and maintenance of gaze as risk factors for poorer outcomes of chronic pain. Participants who displayed prolonged vigilance towards injury images during trials of an initial dot-probe task were more likely than cohorts who gazed longer at neutral images to show significant elevations in pain and interference from pain at a 6-month follow-up. Critically, gaze bias effects were significant, independent of other significant baseline influences (ie, initial pain intensity or interference levels, pain catastrophizing, age, and education) and analysis strategy. In this light, extensions should assess the reliability of gaze biases as predictors of chronic pain outcomes and consider gaze biases as a focus within AB modification interventions.

Back to Top | Article Outline

Conflict of interest statement

The authors have no conflicts of interest to declare.

Back to Top | Article Outline


This research was supported by grants from the China National Natural Sciences Foundation (#31671142) and a 100 Persons Fellowship to the corresponding author.

Back to Top | Article Outline


[1]. Castellanos EH, Charboneau E, Dietrich MS, Park S, Bradley BP, Mogg K. Obese adults have visual attention bias for food cue images: evidence for altered reward system function. Int J Obes 2009;33:1063–73.
[2]. Crombez G, Van Ryckeghem DML, Eccleston C, Van Damme S. Attentional bias to pain-related information: a meta-analysis. PAIN 2013;154:497–510.
[3]. Eccleston C, Crombez G. Pain demands attention: a cognitive-affective model of the interruptive function of pain. Psych Bull 1999;125:356–66.
[4]. Heathcote LC, Lau JYF, Mueller SC, Eccleston C, Fox E, Bosmans M. Child attention to pain and pain tolerance are dependent upon anxiety and attention control: an eye-tracking study. Eur J Pain 2017;21:250–63.
[5]. Jackson T, Chen H, Iezzi T, Ye M, Chen FG. Prevalence and correlates of chronic pain in a random population study of adults in Chongqing, China. Clin J Pain 2014;30:346–52.
[6]. Jackson T, Su L, Wang Y. Effects of higher versus lower threat contexts on pain-related attention biases: an eye-tracking study. Eur J Pain 2018;22:1113–23.
[7]. Jackson T, Su L, Wang Y. Effects of higher versus lower threat contexts on pain-related visual attention biases: an eye-tracking study of chronic pain. J Pain 2018;19:649–59.
[8]. Lautenbacher S, Huber C, Baum C, Rossaint R, Hochrein S, Heesen M. Attentional avoidance of negative experiences as predictor of postoperative pain ratings and consumption of analgesics. Pain Med 2011;12:645–53.
[9]. Lautenbacher S, Huber C, Kunz M, Parthum A, Weber PG, Griessinger N, Sittl R. Hypervigilance as predictor of postoperative acute pain: its predictive potency compared with experimental pain sensitivity, cortisol reactivity, and affective state. Clin J Pain 2009;25:92–100.
[10]. Lautenbacher S, Huber C, Sch€ofer D, Kunz M, Parthum A, Weber PG, Sittl R. Attentional and emotional mechanisms related to pain as predictors of chronic postoperative pain: a comparison with other psychological and physiological predictors. PAIN 2010;151:722–31.
[11]. MacLeod C, Grafton B, Notebaert L. Anxiety-linked attentional bias: is it reliable? Ann Rev Clin Psychol 2019;15:1–26.
[12]. McNeil DW, Rainwater AJ. Development of the Fear of Pain Questionnaire-III. J Behav Med 1998;21:389–410.
[13]. Meng J, Hu L, Shen L, Yang Z, Chen H, Huang X, Jackson T. Emotional primes modulate the responses to others' pain: an ERP study. Exp Brain Res 2012;220:277–86.
[14]. Meng J, Jackson T, Chen H, Hu L, Yang Z, Su Y, Huang X. Pain perception in the self and observation of others: an ERP investigation. Neuroimage 2013;72:164–73.
[15]. Munafo MR, Stevenson J. Selective processing of threat-related cues in day surgery patients and prediction of post-operative pain. Br J Health Psychol 2003;8:439–49.
[16]. Priebe JA, Messingschlager M, Lautenbacher S. Gaze behaviour when monitoring pain faces: an eye-tracking study. Eur J Pain 2015;19:817–25.
[17]. Radloff LS. The CES-D scale: a self-report depression scale for research in the general population. Appl Psych Meas 1977;1:385–401.
[18]. Schoth DE, Delgado Nunes V, Liossi C. Attentional bias toward pain-related information in chronic pain; a meta-analysis of visual-probe investigations. Clin Psychol Rev 2011;32:13–25.
[19]. Sharpe L, Brookes M, Jones E, Gittins C, Wufong E, Nicholas MK. Threat and fear of pain induces attentional bias to pain words: an eye-tracking study. Eur J Pain 2017;21:385–96.
[20]. Sharpe L, Haggman S, Nicholas MK, Dear B, Refshauge K. Avoidance of affective pain stimuli predicts chronicity in acute low back pain patients. PAIN 2014;155:45–52.
[21]. Skinner IW, Hubscher M, Moseley GL, Lee H, Wand BM, Traeger AC, Gustin SM, McAuley JH. The reliability of eyetracking to assess attentional bias to threatening words in healthy individuals. Behav Res Methods 2017;50:1778–92.
[22]. Smith BH, Penny KI, Purves AM, Munro C, Wilson B, Grimshaw J, Chambers WA, Smith WC. The Chronic Pain Grade questionnaire: validation and reliability in postal research. PAIN 1997;71:141–7.
[23]. Sullivan MJ, Bishop SR, Pivik J. The pain catastrophizing scale: development and validation. Psych Assess 1995;7:524–31.
[24]. Tabachnick B, Fidell LS. Using multivariate statistics. 5th edition. London, United Kingdom: Pearson, 2007.
[25]. Todd J, Sharpe L, Colagiuri B, Khatibi A. The effect of threat on cognitive biases and pain outcomes: an eye-tracking study. Eur J Pain 2016;20:1357–68.
[26]. Todd J, Sharpe L, Johnson A, Nicholson-Perry K, Colagiuri B, Dear BF. Towards a new model of attentional biases in the development, maintenance, and management of pain. PAIN 2015;156:1589–600.
[27]. Todd J, van Ryckeghem DM, Sharpe L, Crombez G. Attentional bias to pain-related information: a meta-analysis of dot-probe studies. Health Psychol Rev 2018;12:419–36.
[28]. Van Ryckeghem DM, Crombez G, Goubert L, De Houwer J, Onraedt T, Van Damme S. The predictive value of attentional bias towards pain-related information in chronic pain patients: a diary study. PAIN 2013;154:468–75.
[29]. Vlaeyen JW, Linton SJ. Fear-avoidance and its consequences in chronic musculoskeletal pain: a state of the art. PAIN 2000;85:317–32.
[30]. Wong WS, Fielding R. Prevalence and characteristics of chronic pain in the general population of Hong Kong. J Pain 2011;12:236–45.
[31]. Yang Z, Jackson T, Gao X, Chen H. Identifying selective visual attention biases related to fear of pain by tracking eye movements in a dot-probe paradigm. PAIN 2012;153:1742–8.
[32]. Yang Z, Jackson T, Chen H. Effects of chronic pain and pain-related fear on orienting and maintenance of attention: an eye movement study. J Pain 2013;14:1148–57.
[33]. Yap JC, Lau J, Chen PP, Gin T, Wong T, Chan I, Chu J, Wong E. Validation of the Chinese pain catastrophizing scale in patients with chronic pain. Pain Med 2008;9:186–95.
[34]. Yen CF, Ko CH, Yen JY, Chang YP, Cheng CP. Multi-dimensional discriminative factors for Internet addiction among adolescents regarding gender and age. Psychiatry Clin Neurosci 2009;63:357–64.

Attention biases; Vigilance; Chronic pain; Longitudinal; Eye-tracking

© 2019 International Association for the Study of Pain