Attention and somatic awareness in chronic pain

1. Introduction

One of the disabling features of chronic pain is that it interferes with mental processes and interrupts ongoing activities. Although anecdotally poor concentration is a common clinical complaint, it has attracted little research interest. ^{Jamison et al. (1988)}showed that chronic pain patients report high levels of disruption in concentration and everyday memory. Using a neuropsychological test battery, ^{Kewman et al. (1991)}demonstrated that chronic pain patients who reported high pain intensity at time of testing also suffered greater cognitive disruption. Using a primary task paradigm with clinical populations, ^{Eccleston (1994)and Eccleston (1995a)}showed that those reporting high-intensity chronic pain suffered significant attentional disruption compared to those reporting low-intensity chronic pain and with pain free controls.

Although pain intensity is implicated in the production of attentional disruption, it is not yet clear which psychological factors moderate or mediate this disruption. ^{Dufton (1989)}, in exploring the relationship between pain intensity and affective distress, reported that the subjective complaint of cognitive failure by chronic pain patients was better explained by emotional factors than by the sensory characteristics of the pain. One possible explanation for the role of affective distress in attentional disruption lies in the tendency of those who report high negative affect to be more hypervigilant to bodily sensations and to report more health complaints (^{Watson and Pennebaker, 1989}). ^{Bacon et al. (1994)}, for example, have reported that chronic low back pain patients with greater somatic awareness (as indexed by the frequency of bodily complaints beyond back pain alone) demonstrate more difficulty in attending away from pain than do those with a lesser somatic awareness. This finding could not be attributed to pain sensation alone as there were no significant differences between the reported intensity and quality of the pain.

This experiment is designed to investigate the role of somatic awareness in moderating the disruption of attention by high intensity chronic pain. Using performance on a demanding mental task as the behavioural indicator (^{McCaul and Malott, 1984; Crombez et al., 1994; Crombez et al., 1996}) this experiment focuses upon the retardation of attentional performance caused by the interference of the competing demands of pain intensity and mental load. The chosen task is a Stroop-like task (^{MacLeod, 1991}) that requires patients to overcome the confusion caused by the simultaneous presentation of conflicting information. This task was chosen specifically because its successful performance requires the employment of central attentional processes (^{Logan, 1985; Norman and Shallice, 1986}) and because of its sensitivity as reported by ^{Morton (1969)}. These processes are also responsible for the access of pain into awareness. The Stroop-like interference task has been shown to be sensitive to the attentional demands of pain (^{Eccleston, 1995a}).

Faced with this stimulus one is required to compare the information on one card with the information on the adjacent card. In one condition subjects are required to process the dominant information on the cards by indicating the card with the larger value of digit (here the correct answer would be to press the key marked `left' as `9' is larger than `4'). In a further condition one is required to process the non-dominant information on the cards by indicating which card has the larger number of digits (here the correct answer would be to press the key marked `right' as `6' is greater than `1'). It is known that performance of this task is affected by the recruitment of extra attention in ignoring the dominant response (^{Windes, 1968; Morton, 1969}). The extra time needed to respond to the non-dominant information is referred to as the Number Numerosity Effect (NNE). Where the NNE can be demonstrated it is known that it is detrimentally affected by pain intensity. Indeed, it has been demonstrated that high intensity chronic pain magnifies the NNE (^{Eccleston, 1994; Eccleston, 1995a}). Here, it is hypothesised that performance as measured by the NNE will be significantly impaired in those who report both high intensity chronic pain and high somatic awareness.

2. Methods

2.1 Subjects

Seventy-two chronic pain patients were recruited from a general pain clinic setting. All patients had chronic pain of a persistent and continuous nature. Excluded were patients with head pain or cancer related pain. Forty-six of these consented to participate. Of the 26 who refused 11 gave practical reasons, seven gave hospital transport as the reason, six gave child care as the main reason, and the remaining three refused on grounds of principle. The mean age of the consenting 46 pain patients was 46.8 years (range 24–69 years). Twenty-one patients were male. The average chronicity of complaint was 6.0 years (range 1–40 years). Patients had undergone an average of 4.3 (range 0–12) invasive treatments. Thirty-two patients complained of back pain, six reported widespread muscle pain, two reported limb pain and two reported pain in all sites. Twelve patients were taking no medication, 11 were taking simple or non-steroidal analgesics, six were taking anti-depressant medication alone, six were taking an opioid medication and 11 were taking a combination of analgesic and anti-depressant or opioid medication.

2.2 Measures

(1) A continuous Visual Analogue Scale (VAS) of intensity comprised of a 100 mm horizontal line anchored at left and right by the words `no pain' and `worst imaginable pain', with the line length extending past the anchorage to control for regression bias (^{Carlsson, 1983}). (2) The Short-Form McGill Pain Questionnaire (SFMPQ) (^{Melzack, 1987}) which includes a five word intensity category scale known as the Present Pain Inventory (PPI). (3) A 101 point Numerical Rating Scale (NRS.101) of intensity, which consists of asking the patient to rate his or her pain intensity on a scale of 0–100 where 0 is `no pain' and 100 is a `worst imaginable pain' (^{Jensen et al., 1986}). (4) An 11-word pain intensity Verbal Rating Scale (VRS) (^{Morley, 1988}). Rather than rely on one measure, all five were administered. In line with the main hypothesis, the Modified Somatic Perceptions Questionnaire (MSPQ) (^{Main, 1983}) was also administered. This 22-item instrument is a measure of the frequency and breadth of diffuse somatic complaints. In contrast with other measures of somatic awareness (^{Barsky and Klerman, 1983}), the MSPQ was developed specifically for use with chronic pain patients. Developed first with anxious patients to differentiate somatic perception from a generalised anxiety response, ^{Main (1983)}then analysed and rejected those items, general to anxious populations, that could lead to misclassification in chronic pain populations. He reported significant test-retest reliability on 40 consecutively referred patients, significant construct validity with 300 patients, significant discriminant validity from depression indices and generalised anxiety measures, and clinical validity for back pain and rheumatological related pain. In addition to this primary instrument other measures of affective distress were used: the Hospital Anxiety and Depression Scale (HAD) (^{Zigmund and Snaith, 1983}), as its name implies, is a brief affective measure designed to detect mood disorders in hospitalised populations. It has good internal consistency for both depression and anxiety, and good reliability against psychiatric judgement, with correlations of 0.70 for depression and 0.74 for anxiety in 50 patients (see ^{Bowling (1991)}for further commentary). Although used with pain populations, ^{Williams (1995)}has noted that the brevity of the HAD is largely due to its lack of cognitive items. Therefore also included is a version of the Zung measure of depression modified for chronic pain patients (^{Main and Waddell, 1991}). The predictive validity of this measure has also been examined for a range of pain patients (^{Main et al., 1992}).

2.3 Apparatus

Patients were seated in front of a Macintosh computer at approximately 60 cm from the screen. All stimuli were presented on the screen, responses were made on the computer keypad and all timings were recorded by the computer. Patients responded to serially presented stimuli. Each run contained 22 stimuli. The first two stimuli were considered practice stimuli and were excluded from analyses. Each stimulus comprised two adjacent cards, approximately 25 mm apart. Each card comprised groups of numbers presented within an outline rectangle measuring 85×60 mm (Fig. 1). The numbers were chosen from a pool of 1–9 (inclusive) and appeared in groups of 1–9 items. Two types of information can be attended to on the cards: the value of the digits displayed (V) which was always consistent within a single card, or the number of digits on a card (N). A constraint was imposed that the N of each card never equalled the V of each card. A further constraint on the presentation of the stimuli was also employed: in all conditions neither the same V nor the same N was allowed to occur consecutively.

2.4 Procedure

After giving consent, all patients completed the instruments described above. The true purpose of the study was not explained. All patients prior to testing were shown examples of the cards. It was further explained that with each card two different types of information can be reported: one can attend to the value of the digit on the card (V) or one can attend to the number of digits on the card (N). Patients were required to perform the task twice. In one condition they were instructed to report the position of the card with the larger value of the digit (V). They reported this by pressing a computer key marked `left' or `right'. In Fig. 1 then the correct answer for this condition would be `left' as `9' is greater than `4'. In the second condition they were instructed to report the position of the card with the larger number of digits (N). Again, this was reported via the keyboard. In Fig. 1 the correct answer for this condition would be `right' as `6' is greater than `1'. The order of conditions was counterbalanced.

3. Results

All 46 patients were entered into the analyses. Less than 3% of errors were made, hence accuracy data were not further analysed. All of the following analyses relate to mean correct reaction time data. Also excluded were any individual responses larger than 2.5 SD from each individual mean (^{Tabachnick and Fidell, 1989}).

All patients were classified as being in high or low intensity pain by using a median split of the VAS and the NRS which are highly correlated (r=0.94). Where these were found to be incongruent, the other pain intensity measures were used to classify the patients as high or low. This information was necessary for four patients. This division left 25 patients in the Low Pain group with a mean pain intensity of 27.9 (SD 8.9) (VAS) and 21 in the High Pain group with a mean pain intensity of 60.5 (SD 12.9) (VAS). There were no significant differences between the high and the low intensity groups in terms of age, gender, chronicity of complaint, number of treatments, and medication use.

The predicted interaction between pain intensity and the NNE was investigated in a 2 (Pain: Low and High Pain) ×2 (Task: Value and Number) ×2(Trial: Block 1 and 2) analysis of variance. This is a mixed design with the first variable as between subject and the further two variables as within subject. There was no significant main effect of Task (F(1,44)=2.27) and no significant main effect of Pain (F(1,44)=2.42). There was a significant main effect of Trial (F(1,44)=7.69, P<0.01). The mean of Block 1 was 970 ms and the mean of Block 2 was 925 ms. There were no significant interactions between the above variables. Where ordinal interactions are predicted it is necessary to explore the simple main effects as hypothesised (^{Bobko, 1986}). Therefore planned comparisons were performed on the interactive effects of Pain and Task as a function of Trial. Fig. 2 shows the mean data for each level of these variables.

Patients reporting low-intensity pain (Low Pain group) did not show a significant NNE at Block 1 (t(24)=0.79) or Block 2 (t(24)=−0.61). For patients reporting high-intensity pain (High Pain group) the NNE approached significance in Block 1 (t(20)=1.68, P=0.06), but was significant in Block 2 (t(20)=2.16, P<0.02). In Block 1 there was no difference in the NNE between the High and the Low Pain group (t(44)=0.54). In Block 2, however, the NNE in the High Pain group was significantly larger than the Low Pain group (t(44)=1.89, P<0.05). As the predicted interactive effect on the NNE was found in Block 2 alone, all further analyses of the putative effects of somatic awareness were then performed on the Block 2 data.

Patients were further classified by using a median split of the MSPQ scores. Twenty-two patients were classified as having a low somatic awareness (mean, 4.3; range, 0–7) and 24 were classified as having a high somatic awareness (mean, 13.4; range, 8–24). There were no significant differences between the high and low somatic awareness group in terms of age, chronicity, number of treatments, and medication use. The only subject variable to approach significance was gender. The high somatic awareness group had a proportionally greater number of women (χ²(1)=3.1, P=0.08). A 2 (Pain: Low and High) ×2 (Somatic Awareness: Low and High) ×2 (Task: Value and Number) mixed analysis of variance was performed on the reaction time data for Block 2. As there was a significant positive correlation between the VAS intensity and the MSPQ (r=0.3, P<0.05), the number of subjects in each cell was unequal (Low Pain and Low Somatic Awareness, n=17; Low Pain and High Somatic Awareness, n=8; High Pain and Low Somatic Awareness, n=5; High Pain and High Somatic Awareness, n=16; χ²(1)=8.93 P<0.005). As recommended by ^{Maxwell and Delaney (1990)}, an unweighted analysis of variance was performed. No main effects or two-way interactions were significant. The three-way interaction of interest between Pain Intensity, Somatic Awareness and Task was also not significant (F(1,42)=1.44, P=0.23). Fig. 3 illustrates this interaction.

It is a common feature of such complex interactions for significant effects to be hidden. Therefore, as recommended by ^{Bobko (1986)and Kirk (1983)}, further planned comparisons were performed on this interaction. Pairwise comparisons of the NNE for each of the four groups revealed that the NNE was only significant for the High Pain/High Somatic Awareness group (t(15)=2.049 P<0.03).

For the Low Somatic Awareness group there was no significant difference in NNE between the Low and High Pain groups (t(20)=0.335), but for the High Somatic Awareness group the difference in NNE between Low and High Pain was significant (t(22)=1.973, P<0.03).

3.1 Further analyses

It has been demonstrated that those reporting high-intensity pain who are highly somatically focused show more behavioural disruption as measured by the NNE. As chronic pain patients are a heterogeneous population, further analyses of available data were performed to explore other relationships between variables.

First, as a large proportion of the sample had been prescribed medication, including benzodiazepines, anti-depressants and opioid medication, this may have affected the pattern of results reported above. Although a plausible hypothesis, we did not obtain any supporting evidence. There were no significant differences in medication use between the four experimental groups of patients (χ²(3)=3.533, P=0 32, NS). Further investigating the use of medication use upon task performance, a medication use 4 (Categories 1–4) × task 2 (Number, Group) analysis of variance of reaction time data was performed. There was no significant effect of medication use upon task performance (F(4,39)=0.80, NS). There was also no significant interaction between type of medication and task (F(4,39)=0.30, NS).

Second, it could be argued that the High Pain/High Somatic Awareness group simply reports more intense pain than the High Pain/Low Somatic Awareness group. To test this hypothesis a 2 (Pain: Low and High Pain) ×2 (Somatic Awareness: Low and High) multivariate analysis of variance was performed using all of the pain intensity measures (VAS, NRS, VRS, PPI) as dependent variables. There was no significant main effect of Somatic Awareness (Rao Form 2(4,39)=0.56). However, the interaction of interest between Pain and Somatic Awareness was significant (Rao Form 2(4,39)=2.60, P=0.05). Further univariate analyses indicated that the significant interaction was due to the VAS measure of intensity. However, contrary to the above expectations, for the High Pain groups the High Somatic Awareness group (mean, 58.1) tended to report less intense pain than the Low Somatic Awareness group (mean, 68.0) (t(19)=−1.547, P=0.07).

Third, it has been argued that the tendency to report somatic complaints (as indexed by the MSPQ) is related to a general state of negative affect (^{Watson and Pennebaker, 1989}).

Therefore, a further multivariate analysis of variance was performed using the measures of affective distress (ZUNG and HAD) as dependent variables. There was no main effect of Pain (Rao Form 2(3,40)=1.29). However, a significant main effect of Somatic Awareness emerged (Rao Form 2(3,40)=4.39, P<0.01). Patients reporting high somatic awareness reported more depressive feelings (modified ZUNG: t(44)=2.53, P<0.05; HAD Depression: t(44)=2.82, P<0.01) and more feelings of anxiety (HAD Anxiety: t(44)=3.84, P<0.001) than patients reporting low somatic awareness. The interaction between Pain and Somatic Awareness approached significance (Rao Form 2(3,40)=2.48, P=0.07). Univariate exploration of this interaction revealed that for the Low Somatic Awareness groups there was no difference for the High and the Low Pain group in the mean score of the modified ZUNG, HAD Anxiety and HAD Depression. However, for the High Somatic Awareness groups patients reporting high-intensity pain reported more depressive feelings (modified ZUNG: t(22)=1.92, P<0.04; HAD Depression: t(22)=1.83, P<0.04) and more feelings of anxiety (HAD Anxiety: t(22)=2.197, P<0.02) than the patients reporting low-intensity pain.

4. Discussion

Using a primary task paradigm, performance on an attention demanding task, as indexed by changes in the NNE, was used as a tool to investigate the relationship between chronic pain intensity and somatic awareness. Patients who reported high pain intensity demonstrated impaired performance in an attentionally demanding task as compared to those in low intensity chronic pain. The NNE was differential over performance; it was significantly pronounced in the late stages of the task (Block 2 only). In a further manipulation it was demonstrated that this effect can be accounted for by somatic awareness. Only those patients who report high somatic awareness and high pain intensity show disruption of attention (a pronounced NNE). Those reporting high-intensity pain and low somatic awareness did not demonstrate disruption of attentional performance. Further analyses show that those reporting high-intensity pain and high somatic awareness had significantly less pain than those reporting high pain intensity and low somatic awareness. Further, it is also shown that those reporting high somatic awareness also show more affective distress on a range of measures. Attentional disruption in chronic pain cannot be accounted for by pain intensity alone. These findings have theoretical and clinical implications.

Theoretically, attention has been implicated in the initiation and maintenance of chronic pain (^{Wall, 1993; Wall, 1994}). Attention here is understood as an active process that controls the selection of information into awareness (^{Baddeley and Weiskrantz, 1993}). It has been proposed that this mechanism can explain heightened somatic complaint. ^{Barsky and Klerman (1983)}, for example, have argued that heightened complaint can be explained by a hypervigilance of attention to somatic sensation; the pronounced somatic focus serves to magnify bodily sensations (see also ^{Cioffi, 1991}). The present data are, however, perhaps better explained as being due to the distraction of attention by pain rather than due to the amplification of pain. This is evidenced by the fact that in the high pain group only those demonstrating high somatic awareness suffered greater disruption of attention, despite the fact that their reported pain intensity was similar. The earlier reported study by ^{Bacon et al. (1994)}also offers support for this idea. It is possible that there are two components to hypervigilance: the easier access of pain into awareness and the amplification of pain sensations (See also ^{Rollman and Lautenbacher, 1993}). Although preliminary, this approach offers guidance for further investigation. Important will be to understand better the parameters that govern the access of pain into awareness and the mechanisms by which pain is given attentional priority. One possibility is to explore further the emotional aspects of pain. Here, those in high pain intensity and high somatic awareness also reported the greatest affective distress. Data from other experiments with a non-clinical sample and experimentally induced pain are broadly in support of the idea that anxiety and threat processes may be implicated in the interaction between pain and somatic awareness (^{Crombez et al., 1997b; Crombez et al., 1997c}).

Clinically, the above study adds to the debate concerning the role of attention in the control of chronic pain (^{Leventhal, 1992; Eccleston, 1995b}). Although attention based cognitive strategies are often reported as simple and effective methods of coping with chronic pain, the parameters that govern their effectiveness are unclear. Although the data do not directly address `distraction' as a coping mechanism, it can be suggested that where there is high pain intensity and a high somatic awareness cognitive coping strategies based on distraction may be difficult to apply (see also ^{Heyneman et al., 1990}). As has been previously argued, high intensity chronic pain seems designed to capture attention (^{Eccleston, 1995a}). Where pain is less intense, however, distraction may be an effective strategy. Here, low intensity pain did not interfere with the cognitive task. In line with previous studies patients were able to adequately perform the task in the presence of pain.

There is a number of limitations inherent in this study. First, this study was primarily conducted within a context of the development of an attentional theory of pain (^{Eccleston and Crombez, 1997 submitted}). Therefore a caveat should be heeded regarding generalising to chronic pain populations from this sample. Due to the design employed some cell sizes are small. Before strong clinical implications can be drawn, these data need to be replicated. Second, elided in the current study is the complexity of somatic awareness. The success of further studies in this area are dependent upon achieving greater conceptual clarity regarding the imbricated ideas of hypervigilance, attentional bias, negative affectivity, somatisation and pain amplification. Third, as nociception is not under experimental control, one never exactly measures pain intensity. The detailed investigation of how pain accesses awareness, its effects upon performance and how performance is recovered will probably require a laboratory analogue with experimentally induced pain (^{Crombez et al., 1997a}).

Acknowledgements

This work was supported by a grant to Dr. Eccleston from the Nuffield foundation, UK and a collaborative grant to Drs. Crombez and Eccleston from the National fund for Scientific Research in Belgium and the British Council in Belgium.