Do patients with chronic pain selectively attend to pain-related information?: preliminary evidence for the mediating role of fear

1. Introduction

Investigators have recently extended the use of paradigms commonly used to assess information processing biases in anxiety disorders (for review, see ^{Logan and Goetsch, 1993}) to a variety of other clinical syndromes, including chronic pain. The most widely used task in the assessment of processing biases has been the modified Stroop (1935) colour-naming paradigm. In this paradigm, subjects are shown a series of cues which differ in emotional valence and are asked to name the colour in which a cue is presented while ignoring its meaning. Delays in colour-naming (i.e., Stroop interference) and, by inference, selective attentional processing, occur when the meaning of a cue attracts the subject's attention despite attempts to ignore it. In general, the results of studies employing modified Stroop tests provide an indication of the environmental cues which are attended to by individuals with specific clinical syndromes. These cues typically include those that are of personal relevance (^{Riemann and McNally, 1995}) and are often related to themes of threat (^{McNally, 1994}). To illustrate, spider phobics exhibit Stroop interference for spider-related, but not neutral, cues (^{Watts et al., 1986}). In addition, these cues may represent appropriate _targets for focused cognitive-behavioural intervention (^{Asmundson and Stein, 1995}) and can be used in conjunction with modified Stroop testing to demonstrate effects of treatment (^{Mattia et al., 1993}).

While there have been several studies to address memory biases for pain-related information (e.g., ^{Pearce et al., 1990; Edwards et al., 1992; Edwards and Pearce, 1994; Pincus et al., 1995}) there has been, to the best of our knowledge, only one published study of selective attentional processing in chronic pain patients. Specifically, in an attempt to test the construct validity of the McGill Pain Questionnaire (MPQ; ^{Melzack, 1975}), ^{Pearce and Morley (1989)}administered a modified Stroop test to a sample of 16 patients with chronic pain and to an equal number of healthy control subjects. _target cues included adjectives from the sensory (e.g., throbbing, pounding, sharp) and affective (e.g., tiring, unbearable, agonising) sub-scales of the MPQ, negative emotion cues (e.g., failure, hopeless, depressed), and control cues for each. Results indicated that pain patients, compared to healthy control subjects, had increased Stroop interference for cues drawn from each category of the MPQ but not for the negative emotion cues.

The results of the ^{Pearce and Morley (1989)}study suggest that patients with chronic pain are characterised by a selective attentional bias toward cues thematically related to the (sensory and affective) expression of pain. These findings have yet to be replicated. Moreover, there is some question as to whether the Stroop paradigm is a pure measure of attentional bias (^{MacLeod, 1991}), or of some other bias of information processing (e.g., preoccupation with word meaning) or motor response (e.g., production of appropriate verbal response). Consequently, interpretation related to biases in attentional processing are, at best, tentative.

The purpose of the present investigation was to further probe the nature of the apparent attentional bias for pain-related cues in patients with chronic pain. In order to do so, we used the dot probe paradigm (^{MacLeod et al., 1986}), a task which overcomes the aforementioned limitations of the modified Stroop paradigm by incorporating a neutral motor response (i.e., key press) to a neutral stimulus (i.e., dot-probe). In general, pairings of _target and control cues are presented on a computer monitor and some, but not all, are followed by a dot-probe which appears with equal probability in the spatial location of either member of the pairing. Subjects are required to read the top cue of the pairing aloud and to respond with a key press as quickly as possible upon detecting the dot-probe. Selective attentional processing is inferred when detection latencies in response to dot-probes that follow in the same spatial location as _target cues are speeded relative to those presented opposite the _target cue (i.e., in the spatial location of the neutral cue). Based on the initial findings of ^{Pearce and Morley (1989)}we hypothesised that patients with chronic pain, but not healthy control subjects, would have speeded response latencies for dot-probes presented in the same spatial location as pain-related cues (e.g., stabbing, cramping, punishing). Since all of our patients had pain stemming from musculoskeletal injury, we also tested the corollary hypothesis that response latencies for probes that followed in the spatial location of injury-related cues (e.g., injury, accident, strain) would be speeded in chronic pain patients relative to control subjects.

2. Method

2.1 Subjects

Nineteen patients (seven women; mean age=36.4±9.7 years; range=21–59 years) were recruited from an interdisciplinary rehabilitation program at Wascana Rehabilitation Centre, Regina, Saskatchewan. All patients were experiencing chronic pain (i.e., pain duration of at least 3 months; ^{IASP, 1986}) related to a musculoskeletal injury. In addition, 22 healthy control subjects (nine women; mean age=36.4±9.8 years; range=20–53 years) were recruited from support staff at the Centre. Control subjects were screened for physical and psychiatric illness and were excluded from the study if they reported current persistent pain, current physical illness, or significant emotional difficulties within the past 6 months. Groups did not differ significantly in age, t(39)=0.000, P=1.00, gender distribution, χ² (1)=0.00, P=0.96, or proportion reporting high school graduation (67% vs. 86%, respectively), χ² (1)=1.98, P=0.16. At the time of testing all subjects had normal or corrected-to-normal vision.

All participants provided informed consent and completed a battery of self-report questionnaires following the dot-probe task. This battery comprised the Beck Depression Inventory (BDI; ^{Beck et al., 1961}), a 21-item questionnaire used to assess symptoms of depression, the McGill Pain Questionnaire – short form (MPQ-SF; ^{Melzack, 1987}), comprising a 15-item adjective checklist and two scales for rating pain intensity, and the State-Trait Anxiety Inventory (STAI; ^{Spielberger et al., 1970}), an inventory comprising two brief questionnaires designed to assess state and trait anxiety. They also completed the Anxiety Sensitivity Index (ASI; ^{Peterson and Reiss, 1992}), a 16-item measure of fear of anxiety-related sensations. Evidence suggests that the ASI is conceptually distinct from trait anxiety (^{McNally, 1989; McNally, 1996}) and that it is associated with (e.g., r=0.48; ^{Asmundson and Norton, 1995}) and predictive of fear of pain (e.g., standardised path coefficient=0.55; ^{Asmundson and Taylor, 1996}). As expected, the chronic pain patients scored significantly higher than the control subjects on all self-report measures (see Table 1).

2.2 Stimuli and stimuli presentation

Thirty-two cues, 16 related to pain (e.g., stabbing) and 16 related to injury (e.g., slip) were used as experimental stimuli in this study (see Table 2). The pain-related cues were selected from the MPQ-SF (^{Melzack, 1987}). The injury-related cues were selected from a larger sample of stimuli on the basis of rank orders of actions or events that denote (re)injury or pain that were provided by six clinicians experienced in treating patients with chronic pain. The pain and injury cues were each paired with a neutral word matched for length and frequency of use in the English language (^{Kucera and Francis, 1967}). In addition, 140 cue pairs comprising neutral words were selected from pairings used in previous dot-probe studies (^{Asmundson and Stein 1994, 1995}) to serve as filler material.

All stimuli were presented on a Compaq LTE Elite 4/40C personal computer. Word pairings were presented in the centre of the computer monitor, separated vertically by approximately 3 cm, for a duration of 500 ms. The dot-probe appeared 25 ms after termination of the word pair display on 48 of the 172 trials. Specifically, each of the 16 pain-neutral and 16 injury-neutral word pairings (as well as 16 neutral-neutral word pairings) were followed by a dot-probe. The probability of both the threat cues and the dot-probes appearing in either the upper or lower spatial location on the computer monitor was equated. As in previous dot-probe studies (^{MacLeod et al., 1986; Asmundson and Stein, 1995; Beck et al., 1992}) this created a need for four orders of word pairings, presented in balanced fashion between groups.

Continuation of the computer program was dependent on a response to the computer keyboard on probed trials. This allowed for measurement of the subject's latency to respond to each dot-probe. Trials without dot-probes were followed by another word pairing after an inter-stimulus interval of 1 s.

2.3 Procedure

Subjects were tested individually in an isolated laboratory room. After being seated the subject was informed of the nature of the task by computer-generated instructions. Instructions indicated that: (a) word pairs, separated vertically on the computer monitor, would be presented in series; (b) the top word of the word pairing was to be read aloud; (c) some, but not all, word pairings would be immediately followed by a small dot appearing in the same spatial location as either the top or the bottom word; and (d) upon seeing the dot, the subject was to respond as quickly as possible by pressing the spacebar of the computer keyboard.

2.4 Design and analysis

The design of this study was such that it included one between-group and three within-group factors. The between group factor was Group (chronic pain versus healthy control) and the within-group factors were _target Type (pain versus injury), _target Position (upper versus lower), and Probe Position (upper versus lower). The dependent variable was dot-probe detection latency. Data were analysed using repeated measures analysis of variance (ANOVA). A similar, albeit unplanned, analysis was conducted with the chronic patients divided into two groups on the basis of levels of anxiety sensitivity (and, by inference, fear of pain).

3. Results

The planned repeated measures analysis of detection latencies revealed a main effect for Probe Position, F(1,39)=17.81, P<0.000. Responses were speeded when dot-probes appeared in the upper area of the computer monitor, where subjects' attention was initially allocated, compared to those presented in the lower area (542±131 ms vs. 585±148 ms). There was also a significant main effect for Group, F(1,39)=9.46, P<0.004, with chronic pain patients exhibiting slower overall detection latencies than control subjects (621±164 ms vs. 513±91 ms). As discussed elsewhere (^{MacLeod et al., 1986}), the interpretation of allocation of attention to specific stimuli (i.e., attentional bias) requires, at minimum, a significant interaction between Group, _target Position and Probe Position. _target Type is expected to enter the interaction if specific cues (e.g., pain- versus injury-related cues) selectively capture attention. In the present investigation, there were no theoretically significant interactions between any of the factors. For illustrative purposes, probe detection means and standard deviations for the non-significant Group×_target Position×Probe Position interaction are shown in Table 3. Since our sample of chronic pain patients was more depressed than the control subjects, as assessed with the BDI (^{Beck et al., 1961}), we reanalysed the data with BDI scores as a covariate. Results of this reanalysis did not differ appreciably from those reported above and, therefore, are not reported here. This pattern of results is not surprising given that patients with a primary diagnosis of depression do not show attentional biases for emotionally valenced words presented in the dot-probe task used here (^{MacLeod et al., 1986}). In general, these results represent a failure to replicate evidence of an attentional bias in patients with chronic pain.

In an unplanned analysis we explored the possibility that attentional biases for pain- and injury-related cues are mediated by fear of pain as opposed to chronic pain per se. We did not have a specific measure of fear of pain; however, as discussed above, previous research has shown that anxiety sensitivity is a strong predictor of fear of pain in patients with chronic pain (^{Asmundson and Norton, 1995; Asmundson and Taylor, 1996}). Consequently, we divided the sample of chronic pain patients into groups with high (n=9; mean ASI=26.2±8.7) and low (n=10; mean ASI=9.8±5.6) levels of anxiety sensitivity using a median-spilt of scores from the ASI, assuming that these groupings were also reflective of fear of pain.

Results of the repeated measures analysis of variance indicated, as expected, a significant main effect of Probe Position, F(1,17)=12.36, P<0.005, with speeded response latencies to probes appearing in the upper area relative to those in the lower area (590±157 vs. 654±170). This main effect was modified by a significant Group×_target Position×Probe Position interaction, F(1,17)=5.07, P<0.05. To interpret this interaction we repeated the analysis for each group alone while giving consideration to the same three within-groups factors used in the initial analysis. For the high anxiety sensitivity (high fear of pain) group there were no significant interactions, all Fs<1. For the low anxiety sensitivity (low fear of pain) group, however, there was a significant _target Type×_target Position×Probe Position interaction, F(1,9)=9.08, P<0.05. Further detailed analysis revealed a significant _target Position×Probe Position interaction for the pain-related cues, F(1,9)=26.7, P<0.001. Responses to probes in the upper area were speeded when preceded by a pain-related cue presented in the lower area (630±169 ms) rather than in the upper area (675±202 ms), and responses to probes in the lower area were speeded when preceded by a pain-related cue in the upper area (655±191 ms) rather than in the lower area (708±173 ms). Responses to injury words (see Table 4) were similar across conditions, F<1.

This pattern of results suggests that patients with low fear of pain shift attention away from pain-related words.

4. Discussion

Pearce and Morley's (1989) modified Stroop results suggest that patients with chronic pain selectively attend to cues related to both sensory and affective dimensions of pain. The present study was an attempt to replicate and extend this finding through use of a paradigm (i.e., the dot-probe paradigm) designed to overcome some of the limitations of the modified Stroop paradigm as it relates to inferences of attentional bias. Contrary to our expectations, patients with chronic pain did not differ from control subjects in their pattern of responsivity to dot-probes presented following pain- and injury-related cues. Indeed, with the exception of two uninteresting main effects, there were no significant effects or interactions. This pattern of results represents a failure to replicate Pearce and Morley's (1989) finding of an attentional bias for pain-related cues in patients with chronic pain.

Differences in patient characteristics between the ^{Pearce and Morley (1989)}and present study may have played a role in our failure to replicate. Although ^{Pearce and Morley (1989)}indicate that their patients did not have a life-threatening illness, they did not specify the nature of the pain suffered (e.g., low back pain, headache, menstrual pain) beyond that it was chronic. It is possible that groups of patients with different foci of chronic pain differ in their propensity to selectively attend to pain-related stimuli. Indeed, given the present results in our sample of musculoskeletal pain patients, this may represent an empirical question in need of evaluation.

Alternatively, the findings obtained using the Stroop paradigm (^{Pearce and Morley, 1989}) may have been mediated by some factor unrelated to attentional processing (e.g., rumination over word meaning, response bias). Through incorporation of a neutral response (key press) to a neutral stimulus (dot-probe), the dot-probe paradigm has the advantages of allowing for the direct assessment of allocation of visual attention and, thereby, eliminating interpretations of results that are associated with stages of information processing unrelated to attention and to production of responses. Thus, failure to find significant between-group differences in the pattern of response to pain- and injury-related _target cues presented in the dot-probe paradigm may simply indicate that patients with chronic pain do not attend selectively to these _target cues. That is, it is plausible that the results of ^{Pearce and Morley (1989)}may represent differences between patients and control subjects, not in attention, but either at some other level of information processing or in the production of their responses.

Results from of our analysis of probe detection latency between patients with high and low levels of anxiety sensitivity are intriguing and, contrary to the above discussion, suggest that some chronic pain patients may selectively attend to pain-related stimuli. When patients were divided on the basis of scores on the ASI (^{Peterson and Reiss, 1992}) and, by inference, fear of pain (^{Asmundson and Norton, 1995; Asmundson and Taylor, 1996}), those with high fear responded in a similar fashion to all stimuli (i.e., they did not exhibit a tendency to selectively attend to pain-related stimuli). On the other hand, patients with low fear had speeded responses to probes in the upper area when preceded by a pain-related cue presented in the lower area (630±169 ms) rather than in the upper area (675±202 ms), and speeded responses to probes in the lower area when preceded by a pain-related cue in the upper area (655±191 ms) rather than in the lower area (708±173 ms). That is, these patients shifted attention away from stimuli related to sensory and affective dimensions of pain. Thus, in situations thematically related to pain, the operation of the information processing system may differ substantially depending on a patient's trait predisposition to fear (or not fear) pain-related stimuli.

A tendency to shift attention away from environmental cues that are related pain may have implications for the way one behaves in their environment. For example, this style of information processing may be related to coping strategies characterised by cognitive avoidance, in which one distracts from, or ignores, pain (for review, see ^{Jensen et al., 1991}). Moreover, a failure to process pain-related cues beyond the level of attention may lead to an underestimation of their impact or importance. The process of shifting attention away from pain cues, combined with underestimates of their significance, may be related to (or, may even promote) participation in inappropriate activity (e.g., overexertion), potential reinjury, and continued pain in patients with low fear (also see ^{McCracken et al., 1993}). On the other hand, based on a review of predictions made following induction of acute pain, ^{Arntz (1996)}has suggested that underprediction of pain may increase fear, caution, avoidance and preparation to fight or flee and is, thereby, important to survival. These mixed conclusions indicate that investigations geared toward reconciling the relationship between attentional biases, fear of pain, and pain prediction in both acute and chronic pain are needed to advance our understanding in this area. We are currently conducting studies to address some of these issues, including one to confirm that the conclusions of the present study will hold when a direct measure of fear of pain is employed.

In summary, the results of the present investigation do not support previous findings of a general attentional bias for pain-related cues in patients with chronic pain (^{Pearce and Morley, 1989}). Our hypothesis that these patients would selectively attend to injury-related cues was also unsupported. We did find, however, that chronic pain patients with low anxiety sensitivity (low fear of pain) exhibit a tendency to shift attention away from pain-related environmental stimuli. The direct assessment of visual attention employed in this study ensures that these findings were not the result of cognitive or response processes unrelated to attention. Notwithstanding, the clinical significance of the observed attentional bias remains to be determined.

Acknowledgements

Preparation of this manuscript was supported, in part, by a grant to the first author from the Saskatchewan Health Services Utilization and Research Commission. The authors would like to thank Marilee Allerdings and Carla Hale for assistance with data collection.