Measures of anxiety: is there a difference in their ability to predict functioning at three-month follow-up among pain patients?

Transcription

1 European Journal of Pain 8 (2004) Measures of anxiety: is there a difference in their ability to predict functioning at three-month follow-up among pain patients? Heather D. Hadjistavropoulos a,b, *, Gordon J.G. Asmundson a,b, Kristine M. Kowalyk a,b a Department of Psychology, Faculty of Kinesiology and Health Studies, University of Regina, Regina, Saskatchewan, Canada S4S 0A2 b Research and Performance Support, Regina QuÕAppelle, Health Region, Regina, Saskatchewan, Canada S4S 0A6 Received 4 November 2002; accepted 22 April 2003 Abstract Independent investigators have found that pain is related to health anxiety, trait anxiety, pain-related anxiety, and anxiety sensitivity. To date, the relationship among these anxiety-related constructs has not been studied directly and little is known about their relative ability to predict adjustment to pain over time. This paper presents longitudinal data from measures given to 227 musculoskeletal pain patients. Patients were asked at the time of their first visit (T1) to a physiotherapy clinic to complete a questionnaire package including measures of these different forms of anxiety as well as pain severity, disability, negative affect, and perceived control. Approximately 3 months later (T2), 50% of patients responded to these same questionnaires. Results showed that correlations among the anxiety measures at T1 ranged from 0.35 to Using multiple regression analyses, measures of T1 anxiety were each examined for their ability to predict unique variance in disability, negative affect, and perceptions of control measured at T1 and T2. At T1, after controlling for pain severity and other measures of anxiety, pain-related anxiety uniquely predicted both disability and negative affect, trait anxiety uniquely predicted negative affect and perceptions of control, and anxiety sensitivity uniquely predicted negative affect. At T2, after controlling for pain severity, other measures of anxiety and each respective measure of functioning at T1, health anxiety uniquely predicted disability and negative affect, although anxiety sensitivity also uniquely contributed to the prediction of negative affect. It is concluded that the importance of various forms of anxiety is dependent on the timeframe and outcome examined. Clinical implications of the findings as well as directions for future research are discussed. Ó 2003 European Federation of Chapters of the International Association for the Study of Pain. Published by Elsevier Ltd. All rights reserved. Keywords: Health anxiety; Anxiety sensitivity; Trait anxiety; Pain-related anxiety; Longitudinal design; Pain assessment; Disability 1. Introduction Individual differences in the experience of pain are common and widely documented in the pain research literature (e.g., Turk and Rudy, 1987, 1988). Thus, despite experiencing the same painful stimulus, individuals experience pain in different ways. One explanation for these differences is that both psychological and social factors play a role in the experience of pain (Burton et al., 1995; Kerns and Jacob, 1992; Linton and Skevington, 1999). Anxiety is one factor that has consistently been shown to influence perception and adjustment to * Corresponding author. Tel.: ; fax: address: hadjista@uregina.ca (H.D. Hadjistavropoulos). pain. For instance, various anxiety constructs such as pain-related anxiety (McCracken et al., 1992), health anxiety (Kellner, 1986), trait anxiety (Spielberger et al., 1970), and anxiety sensitivity (Reiss et al., 1986), have been shown to be related to the experience of pain. These constructs are briefly discussed below. A burgeoning area of investigation in the study of chronic pain is concerned with fears of pain (for recent reviews see Asmundson et al., 1999a; Vlaeyen and Linton, 2000). The Pain Anxiety Symptom Scale (PASS), for instance, was developed to measure fear of pain along four dimensions, including fearful interpretations, avoidance and escape, physiological responses, and symptoms of cognitive interference (McCracken et al., 1992). Investigators using this measure have shown that, after controlling for pain severity, high scores on the /$30 Ó 2003 European Federation of Chapters of the International Association for the Study of Pain. Published by Elsevier Ltd. All rights reserved. doi: /s (03)

2 2 H.D. Hadjistavropoulos et al. / European Journal of Pain 8 (2004) 1 11 PASS are associated with higher levels of affective distress, dysfunction, and disability (e.g., Asmundson et al., 1997; Crombez et al., 1999; McCracken et al., 1992; McCracken et al., 1993). Avoidance beliefs, one dimension of pain-related anxiety measured by the Fear- Avoidance Beliefs Questionnaire (FABQ), have similarly been found to be associated with disability, poor behavioural performance, and also more days of work loss (Crombez et al., 1999; Waddell et al., 1993). Additionally, it has been shown that pain-free individuals scoring high on the FABQ, were twice as likely to experience an episode of pain over the course of the following year (Linton et al., 2000). Rather than focussing specifically on anxiety about pain, others have studied the relationship between pain adjustment and anxiety about health. Health anxiety refers to a continuum, with mild to no concern about bodily sensations at one end and preoccupation with and fear of bodily symptoms and conviction in disease at the other (Salkovskis and Warwick, 1986). A recent review of the literature concerning health anxiety and pain shows that health anxiety is elevated among pain samples as compared to controls (Hadjistavropoulos et al., 2000a). Furthermore, there is convincing evidence to suggest that health anxiety affects a patientõs physical, behavioural, and cognitive response to pain. In particular, health anxiety is related to increased pain, somatic sensations, avoidance, reassurance seeking, as well as greater attention to symptoms and dysfunctional cognitions regarding pain (see Hadjistavropoulos et al., 2000a). Still others researching anxiety and pain have focussed more generally on trait anxiety. Trait negative affectivity, also referred to as neuroticism, general maladjustment, and trait anxiety (Watson and Clark, 1984), has been defined as a broad dimension of individual differences in the tendency to experience negative, distressing emotions... (Costa and McCrae, 1987, p. 301). One instrument used to measure this construct is the State-Trait Anxiety Inventory (STAI; Spielberger et al., 1970). In a factor analysis of several outcome measures used to assess the impact of recurrent headaches, it was found that the STAI loaded most strongly on an affective distress factor (Holroyd et al., 1999). This result lends support to the use of the STAI as an indicator of the amount of affective distress experienced by those suffering recurrent pain. Researchers have found that trait anxiety correlates with pain severity and disability among those with musculoskeletal pain (McCracken et al., 1996). Likewise, in a sample of patients with workrelated muscular pain (for the most part low back pain) or fibromyalgia, those who showed high trait anxiety reported increased catastrophizing, and lower ability to control and reduce pain (Hallberg and Carlsson, 1998). One other related construct that is gaining increasing attention in the area of pain is anxiety sensitivity. Anxiety sensitivity measures fear of anxiety-related bodily sensations, which is thought to arise from beliefs that these sensations will result in harmful somatic, social, or psychological consequences (Reiss and McNally, 1985). Regardless of pain severity, patients with high anxiety sensitivity scores respond with fear and avoidance to a number of stimuli, including pain (Asmundson and Taylor, 1996), and appear to be at risk of increased pain behaviour, emotional distress, and lowered perceptions of self-control (Asmundson and Norton, 1995; Asmundson et al., 1999b; Norton et al., 1995). Components of functional status (i.e., vitality, mental health, and social functioning) among chronic pain patients have also been found to be related to anxiety sensitivity irrespective of pain severity (Plehn et al., 1998). Exploring more specifically the underlying component fears represented by the anxiety sensitivity construct, research by Asmundson et al. (2000) suggests that cognitive and emotional fears may have the strongest link to the experience of pain, with fear of somatic sensations being less relevant. Two pain induction studies have also been conducted (i.e., Keogh and Birkby, 1999; Keogh and Mansoor, 2001) and, although the results of these studies are preliminary, their findings suggest that high anxiety sensitivity influences oneõs pain experience as opposed to its influence working in the opposite direction (also see Asmundson, 2001). Limited research attention has been directed toward understanding the relative importance and distinctiveness of these anxiety constructs in understanding adjustment to pain. Previous research has shown that anxiety about pain provides better information about pain, disability, and avoidance than more generalized anxiety (e.g., trait anxiety, as measured by the STAI; McCracken et al., 1996; McCracken et al., 1992). Further, trait anxiety and health anxiety have also been shown to be moderately correlated (Hitchcock and Mathews, 1992), yet distinct constructs. High health anxiety is specifically associated with reassurance seeking and seeking out medical attention, as well as with catastrophizing even after controlling for trait anxiety (Hitchcock and Mathews, 1992; Watson and Pennebaker, 1989). With respect to anxiety sensitivity, researchers have found that, after controlling for pain severity, individuals obtaining high scores on a measure of anxiety sensitivity also experience high levels of painrelated anxiety, trait anxiety, negative affect, and low control (Asmundson and Norton, 1995; Asmundson et al., 1999b). While the above research represents a positive beginning, further research is needed to examine the anxiety constructs collectively in relation to pain adjustment. While it is clear that anxiety, in one form or another, is concurrently related to the pain experience, the extent to which the various anxiety-related constructs are predictive of long-term adjustment to pain remains unclear.

3 H.D. Hadjistavropoulos et al. / European Journal of Pain 8 (2004) To date, McCracken and Gross (1998) have conducted one of the only prospective studies of pain-related anxiety and its association with chronic pain. Specifically, in this study, pretreatment and posttreatment measures of pain-related anxiety were obtained from a sample of chronic low back pain patients enrolled in a 3-week multidisciplinary, functional treatment program. At posttreatment, lower pain-related anxiety was associated with increased activity and lower ratings of depression, pain severity, disability, and affective distress. The purpose of the present study is to address the paucity of prospective studies regarding anxiety and adjustment to pain by examining the relationship among anxiety constructs and their relative importance individually, and in combination, in predicting adjustment to pain concurrently and over time. Taylor (1995), as well as Zinbarg and colleagues (1997), have suggested that there is a hierarchical structure to fears, with trait anxiety at one of the higher levels and more specific fears comprising the lower order factors. Our intent here was to determine the relative importance of these anxiety constructs, regardless of their order in the hierarchy, in predicting concurrent and long-term adjustment to pain. Based on the findings of the research reviewed herein, it is expected that each of the anxiety constructs will explain unique variance in pain adjustment as well as jointly account for explained shared variance. Given the exploratory nature of this research, more specific hypotheses are not warranted. However, we speculate that because trait anxiety may be a higher order factor, the amount of unique variance explained by this construct will be minimal as a result of shared variance. 2. Method 2.1. Participants and procedure Consecutive pain patients referred by their physician to a physiotherapy clinic over a one and a half-year period were invited to participate in a study of how pain and response to pain progress over time. Those who consented, completed a questionnaire package. The package was completed at the clinic in approximately 30 minutes. Three months after the initial clinic visit, the package was re-administered by mail. During the threemonth period, patients underwent physiotherapy that was tailored to their specific condition. This typically involved a combination of exercises and pain relief modalities (e.g., ice, heat, transcutaneous muscle stimulation). Detailed data were not recorded on the nature of treatment since the interest of the study was on anxiety as a predictor of functioning over time. During the initial clinic visit (T1), a total of 227 patients agreed to participate. The mean age of the sample was 44.1 years (SD ¼ 16:2), 57.9% were female, and 63.6% were married. The average education was 13.3 years (SD ¼ 2:6). Only 9.1% were receiving compensation for their injury. The mean pain duration reported by the sample was 25.7 months (SD ¼ 52:2). At T1, 54.8% of participants were classified as having acute pain (pain less than three months) compared to 45.2% who were classified as having pain lasting more than 3 months. The most commonly reported primary pain sites, coded using the International Association for the Study of Pain (1986) primary pain site coding system, included the upper limbs (35.5%), followed by the lower limbs (32.2%), and the back (15.7%). Other reported areas included the neck (5.8%), groin (5.8%), chest (1.7%), and head (0.8%). Only 2.5% of the sample reported primary pain located in multiple areas. Approximately 46.6% reported experiencing pain on a constant versus sporadic basis. At 3-month follow-up (T2), approximately half of the patients completed the questionnaire package again. Thus, the final sample for the present study comprised 121 pain patients. The final sample did not differ significantly from the initial sample of 227 on any background, pain-adjustment, or anxiety variables, with the exception of age, F ð1; 225Þ ¼9:05, p < 0:003, and education, F ð1; 225Þ ¼10:37, p < 0:003. The initial sample was slightly younger (M ¼ 38:0; SD ¼ 14:1) and slightly less educated (M ¼ 12:3; SD ¼ 2:2) Measures Background variables Participants completed a background information questionnaire to assess age, sex, marital status, financial compensation, and education. Date of present injury and course of pain were also queried Pain-related adjustment and control Section I of the Multidimensional Pain Inventory (MPI; Kerns et al., 1985) was administered to assess pain severity, disability, negative affect, social support, and perceived control. Participants responded to 28 questions on a numerical scale from 0 to 6 anchored by various descriptive phrases (e.g., no pain/extreme pain, not at all worried/extremely worried, no control/extreme control). The scale has been shown to have good reliability and validity with heterogeneous samples of chronic pain patients (Kerns and Jacob, 1992). The social support questions of Section I were not of interest conceptually and, therefore, were not reported in the present study Pain anxiety symptom scale The Pain Anxiety Symptom Scale (PASS; McCracken et al., 1992) is a 40-item self-report scale that measures four components of fear of pain, including cognitive anxiety, fearful appraisals of pain, escape and

4 4 H.D. Hadjistavropoulos et al. / European Journal of Pain 8 (2004) 1 11 avoidance, and physiological anxiety. Items are rated for frequency of occurrence on a 6-point scale, from 0 (never) to 5 (always), and yield scores on the aforementioned components as well as a total score. The PASS has been found to have both good test retest reliability (McCracken et al., 1993) and validity (Burns et al., 2000; McCracken et al., 1992). As other researchers have done (e.g., McCracken and Gross, 1998), the total PASS score was used for the purposes of the present study Illness attitudes scale The Illness Attitudes Scale (IAS; Kellner, 1986; Kellner et al., 1987) consists of 27 items, rated on a 0 (no) to4(most of the time) scale, examining fears, beliefs, and attitudes related to health anxiety. Nine subscales were initially identified, but a recent factor analysis of the scale suggests that a hierarchical model is more appropriate, with four lower order factors loading on a single higher order factor (i.e., fear of illness, effects of symptoms, treatment experiences, and disease conviction; Hadjistavropoulos et al., 1999). To obtain a score of overall health anxiety, the items were summed. The total score has been used in previous research examining health anxiety (Hadjistavropoulos et al., 1998; Hitchcock and Mathews, 1992). The IAS has been found to be both reliable and valid (Hadjistavropoulos et al., 2000b) Anxiety sensitivity The Anxiety Sensitivity Index (ASI; Reiss et al., 1986) was included to measure fear of anxiety signs and symptoms. Three lower order components of the anxiety sensitivity construct include fear of social concerns, fear of physical catastrophe, and fear of mental incapacitation (Zinbarg et al., 1997; Zinbarg et al., 1999). The questionnaire includes 16 items rated on a 1 (very little) to 4 (very much) scale, which are summed to obtain a total score. The ASI has been demonstrated to be reliable and valid (McNally, 1996; Peterson and Reiss, 1992) Trait anxiety The trait form of the State-Trait Anxiety Inventory (STAI-T; Spielberger et al., 1970) was used to examine trait anxiety. Self-ratings regarding how participants generally feel are made on a 1 (almost never) to4(almost always) scale. The items are summed to obtain a total score. The measure has been found to be both reliable and valid (Spielberger et al., 1993) Statistical analyses Multiple regression analyses were performed for each of the following dimensions: MPI disability, MPI negative affect, and MPI perceived control. Separate analyses were performed to determine the predictability of each MPI variable at T1 and T2 using T1 measures of the anxiety constructs. A forced-entry method was used to examine the predictive strength of each of the anxietyrelated constructs: (1) after controlling for MPI pain severity at T1, when predicting the various other MPI variables at T1; and (2) after controlling for MPI pain severity at T1 and each respective measure of functioning at T1, when predicting the MPI variables at T2. For example, in the regression analysis involving MPI disability at T2, first MPI pain severity at T1 and MPI disability at T1 were entered, followed by the four anxiety constructs. This allowed for examination of the unique and independent contribution of each anxiety construct within a set. The analyses were also carried out controlling for pain duration in step 1 along with pain severity. This did not change the results of the analyses predicting adjustment variables at either T1 or T2. 3. Results Means and standard deviations of the variables measured in the study are presented in Table 1. ASI scores were within the moderate range (M ¼ 16:5, SD ¼ 10:1) as outlined by Asmundson et al. (1999a) and fell within the mean range of other samples of musculoskeletal pain patients (M ¼ 15:0 17:6; Asmundson et al., 1998; Asmundson et al., 1997; Asmundson and Norton, 1995; Asmundson and Taylor, 1996). The mean IAS score (M ¼ 32:5, SD ¼ 13:2) closely matched that of another sample of musculoskeletal injury patients (M ¼ 35:2, SD ¼ 10:5; Hadjistavropoulos et al., 2000b) with scores falling at least one standard deviation above the mean IAS score of a sample of controls (M ¼ 26:2, SD ¼ 7:9; Hadjistavropoulos, 1995). The mean PASS score (M ¼ 58:3, SD ¼ 25:9) was lower than was previously found in a sample of chronic back pain patients (M ¼ 87:4, SD ¼ 31:1; McCracken and Gross, 1998) and another sample of patients referred to a pain clinic (M ¼ 94:2, SD ¼ 39:2; McCracken et al., 1993). This may be a reflection of the fact that the present sample was a community sample and not attending interdisciplinary treatment. The mean STAI-T score (M ¼ 41:1, SD ¼ 8:5) was similar to the mean scores from a sample of first-year cadet university students and nursing students (M ¼ 38:4, SD ¼ 11:2; Schmidt et al., 1999; M ¼ 38:23, SD ¼ 7:49; Moffett et al., 1993). Repeated measures analyses indicated that there were no significant changes in the anxiety variables from T1 to T2 (ASI, F ð1; 113Þ ¼0:07, p ¼ 0:79; IAS, F ð1; 113Þ ¼0:39, p ¼ 0:53; PASS, F ð1; 113Þ ¼3:04, p ¼ 0:08; and STAI-T, F ð1; 113Þ ¼2:60, p ¼ 0:11). Correlations among the constructs were moderate ranging from 0.35 to 0.56 (see Table 2). At T1, means for MPI pain severity and disability were within the moderate range (M ¼ 3:1, SD ¼ 1:3;

5 H.D. Hadjistavropoulos et al. / European Journal of Pain 8 (2004) Table 1 Means and standard deviations of the study variables Variable Mean Standard deviation MPI pain severity Time Time MPI disability Time Time MPI negative effect Time Time MPI perceived control Time Time Health anxiety (IAS) Time Time Trait anxiety (STAI-T) Time Time Pain-related anxiety (PASS) Time Time Anxiety sensitivity (ASI) Time Time Note. MPI, multidimensional pain inventory; IAS, illness attitude scale; STAI-T, state-trait anxiety inventory-trait form; PASS, pain anxiety symptom scale; ASI, anxiety sensitivity index. Table 2 Correlations among anxiety measures (N ¼ 227) Measures STAI-T PASS ASI IAS State-trait anxiety inventory (STAI-T) Pain anxiety symptom scale (PASS) Anxiety sensitivity 0.56 index (ASI) Illness attitudes scale (IAS) Note. Trait anxiety was measured by the state-trait anxiety inventory (STAI-T); pain-related anxiety was measured by the pain anxiety symptom scale (PASS); anxiety sensitivity was measured by the anxiety sensitivity index (ASI); and health anxiety was measured by the illness attitudes scale (IAS). All correlations were significant at the p < 0:01 level. M ¼ 2:8, SD ¼ 1:4; pain severity and disability, respectively). At T2, means for MPI pain severity and disability scales moved to within the low to moderate range (M ¼ 1:9, SD ¼ 1:5; M ¼ 1:8, SD ¼ 1:4; respectively). Approximately 25% of patients reported a score of 1 or lower on the MPI pain severity scale, and approximately 30% of patients reported a scored of 1 or lower on the MPI disability scale. Mean MPI negative affect was within the moderate range at T1 (M ¼ 2:2, SD ¼ 1:2), increasing slightly but remaining in the moderate range at T2 (M ¼ 2:6, SD ¼ 0:9). Mean MPI perceived control fell within the high range of the scale at T1 (M ¼ 3:8, SD ¼ 1:1) and stayed in the high range, but increased slightly at T2 (M ¼ 4:2, SD ¼ 1:1; possible scale range from 0 to 6). Repeated measures analyses indicated that for each of these variables there was a significant change in score from T1 to T2 (pain severity, F ð1; 116Þ ¼52:38, p < 0:001; disability, F ð1; 116Þ ¼34:32, p < 0:001; negative affect, F ð1; 116Þ ¼26:15, p < 0:001; and perceived control, F ð1; 116Þ ¼3:87, p ¼ 0:05) Predicting disability To determine the amount of variance in T1 disability explained by the anxiety constructs measured at T1, a multiple regression analysis was performed, controlling for pain severity at T1 (b ¼ 0:66, t ¼ 12:85, p < 0:001; see Table 3). Entered on the first step, pain severity accounted for 43% of the variance in disability (F ð1; 217Þ ¼ 181:99, p < 0:001). Including pain severity with the various anxiety constructs (second step), 50% of the variance in disability was explained (F ð5; 213Þ ¼42:28, p < 0:001). The four anxiety constructs explained an additional 7% of the variance in disability (DR 2 ¼ 0:07, p < 0:001) above pain severity. Inspection of the squared semi-partial correlations indicated that the PASS was the only anxiety construct to significantly contribute to the prediction of disability at T1 (b ¼ 0:29, t ¼ 4:44, p < 0:001), explaining approximately 5% of the unique variance in disability. To assess the contributions of the anxiety constructs measured at T1 to the prediction of T2 disability, another regression analysis was performed. To control for T1 pain severity and disability, these variables were entered on step one. Together, initial pain severity and disability explained 29% of the variance in T2 disability (F ð2; 114Þ ¼23:52, p < 0:001; see Table 4), although, only initial disability significantly contributed to the prediction of future disability (b ¼ 0:44, t ¼ 4:37, p < 0:001; sr 2 ¼ 0:12). On step two, having controlled for pain severity and disability at T1, STAI-T, ASI, IAS, and PASS at T1 collectively predicted a significant additional 7% of the variance in T2 disability (DR 2 ¼ 0:07, p < 0:05). Closer inspection revealed that only the IAS explained a significant portion of the variance in disability (b ¼ 0:21, t ¼ 2:29, p ¼ 0:024), with examination of the squared semi-partial correlation further revealing that it accounted for approximately 3% of the unique variance at T2. In total, 36% of the variance in T2 disability was explained when T1 pain severity and disability were entered into the regression analysis with the anxiety constructs (F ð6; 110Þ ¼10:25, p < 0:001).

6 6 H.D. Hadjistavropoulos et al. / European Journal of Pain 8 (2004) 1 11 Table 3 Summary of multiple regression analyses for the prediction of pain severity, disability, negative affect, and perceived control at time 1 (N ¼ 190) Variable B SE B b t sr 2 R 2 DR 2 Disability PS *** *** PS *** *** STAI-T ) )0.105 ) PASS *** ASI ) )0.118 ) IAS Negative affect PS *** *** PS *** *** STAI-T ** PASS * ASI *** IAS Perceived control PS ) )0.349 )5.47*** *** PS ) )0.254 )4.16*** *** STAI-T ) )0.298 )4.38*** PASS ) )0.110 ) ASI ) )0.007 ) IAS ) )0.070 ) Note. PS, MPI pain severity at time 1; STAI-T, state-trait anxiety inventory; PASS, pain anxiety symptom scale; ASI, anxiety sensitivity index; IAS, illness attitudes scale. *p < 0:05. **p < 0:01. ***p < 0: Predicting negative affect To predict T1 negative affect using the four anxiety constructs also measured at T1, a multiple regression analysis was performed controlling for pain severity (b ¼ 0:56, t ¼ 9:85, p < 0:001; see Table 3). Pain severity was entered on the first step of the analysis and was found to account for 31% of the variance in negative affect (F ð1; 215Þ ¼96:96, p < 0:001). Entered on step 2, the four anxiety constructs significantly contributed an additional 24% to the prediction of negative affect (DR 2 ¼ 0:24, p < 0:001). Fifty-five percent of the variance in negative affect was accounted for by the measures of pain severity and the four anxiety constructs (F ð5; 211Þ ¼50:76, p < 0:001). Examination of the squared semi-partial correlations revealed that the ASI, STAI-T, and PASS all contributed significantly to the prediction of T1 negative affect (b ¼ 0:29, t ¼ 4:94, p < 0:001; b ¼ 0:14, t ¼ 2:62, p < 0:01; b ¼ 0:16, t ¼ 2:56, p < 0:05; respectively). Each of the constructs uniquely contributed 5%, 2%, and 1%, respectively. A multiple regression assessing the various T1 anxiety constructs contributions to the prediction of T2 negative affect was similarly performed. In the first step of the analysis, T1 pain severity and negative affect were entered, accounting for 16% of the variance in T2 negative affect (F ð2; 112Þ ¼10:80, p < 0:001; see Table 4). Only T1 negative affect accounted for a significant portion of the variance explained by these two variables (b ¼ 0:38, t ¼ 3:72, p < 0:001; sr 2 ¼ 0:10). In step 2, the four measures of anxiety collectively accounted for an additional 9% of the variance in T2 negative affect (DR 2 ¼ 0:09, p ¼ 0:013). Individually, only the IAS and ASI explained a significant portion of the variance in T2 negative affect (b ¼ 0:23, t ¼ 2:29, p < 0:05; b ¼ 0:33, t ¼ 3:09, p < 0:01; respectively). Inspection of the squared semi-partial correlations revealed that the IAS accounted for a unique 4% of the variance, and the ASI for a unique 7% of the variance in T2 negative affect. T1 pain severity, negative affect, and the four anxiety constructs, together accounted for 25% of the variance in T2 negative affect (F ð6; 108Þ ¼6:10, p < 0:001) Predicting perceived control Entered on the first step, T1 pain severity explained 12% of the variance in beliefs regarding perceived control at T1 (F ð1; 215Þ ¼29:92, p < 0:001; b ¼ 0:35,

7 H.D. Hadjistavropoulos et al. / European Journal of Pain 8 (2004) Table 4 Summary of multiple regression analyses for the prediction of pain severity, disability, negative affect, and perceived control at time 2 (N ¼ 116) Variable B SE B b t sr 2 R 2 DR 2 Disability (time 2) PS DIS PS DIS STAI-T PASS ASI ) )0.101 ) IAS Negative affect (time 2) PS NA PS ) )0.021 ) NA STAI-T PASS ASI ) )0.326 ) IAS Perceived control (time 2) PS ) )0.115 ) CON PS ) )0.105 ) CON STAI-T ) )0.235 ) PASS ) )0.044 ) ASI IAS ) )0.027 ) Note. PS, MPI pain severity at time 1; STAI-T, state-trait anxiety inventory; PASS, pain anxiety symptom scale; ASI, anxiety sensitivity index; IAS, illness attitudes scale; DIS, disability at time 1; NA, negative affect at time 1; CON, perceived control at time 1. * p < 0:05. ** p < 0:01. *** p < 0:001. t ¼ 5:47, p < 0:001; see Table 3). Entry of pain severity along with the four anxiety constructs on the second step of the analysis explained 28% of the variance in perceived control (F ð5; 211Þ ¼16:17, p < 0:001). The four anxiety constructs on their own accounted for 16% of the variance in the measure of T1 control (DR 2 ¼ 0:16, p < 0:001) beyond pain severity. Examined individually, the squared semi-partial correlations revealed that only the STAI-T contributed significantly to the prediction of T1 perceived control (b ¼ 0:30, t ¼ 4:38, p < 0:001), uniquely accounting for 7% of the variance. To assess the contributions of the anxiety constructs measured at T1 to the prediction of T2 perceived control, a final multiple regression analysis was performed. On the first step, T1 pain severity and perceived control accounted for 14% of the variance in T2 perceived control (F ð2; 112Þ ¼9:10, p < 0:001; see Table 4), although only T1 perceived control explained a significant portion of the explained variance (b ¼ 0:31, t ¼ 3:33, p < 0:01; sr 2 ¼ 0:09). The four measures of anxiety, in conjunction with T1 pain severity and perceived control, explained 20% of the variance at T2 (F ð6; 108Þ ¼4:41, p < 0:001). The addition of the anxiety constructs on the second step did not result in a significant increase in the amount of variance explained (DR 2 ¼ 0:06, p ¼ 0:112). Therefore, the only unique significant predictor of T2 perceived control was the level of perceived control at T1, explaining 9% of the variance. 4. Discussion The purpose of this study was to determine whether the anxiety constructs previously found to be related to pain and pain adjustment (i.e., trait anxiety, anxiety sensitivity, health anxiety, and pain-related anxiety) could be used to predict longitudinally the functioning of patients suffering from pain at 3-month follow-up. This study focused on a community sample as compared

8 8 H.D. Hadjistavropoulos et al. / European Journal of Pain 8 (2004) 1 11 to a chronic pain sample attending interdisciplinary treatment where anxiety is expected to be elevated to a much greater extent. Further, the combined and unique contributions of the various anxiety constructs were of interest. Before discussing the findings with regard to the prediction of future adaptation to chronic pain, we briefly examine the findings regarding the factors that predicted functioning at the time of initial assessment Predicting concurrent functioning Results indicated that each of the various anxiety constructs was related to pain adjustment, but the importance varied depending on the specific measure of adaptation examined. Additionally, the results showed that, even while controlling for other forms of anxiety, each of the anxiety constructs was distinct and contributed uniquely to pain adjustment. The one exception to this was health anxiety, which did not contribute uniquely to concurrent pain adjustment. For instance, pain-related anxiety significantly and uniquely contributed to the prediction of disability at T1. Previous research has demonstrated the association of pain-related anxiety to disability (e.g., Asmundson and Larsen, 2000; Crombez et al., 1999; McCracken et al., 1993; Waddell et al., 1993), but not while controlling for other diverse forms of anxiety. Three of the anxiety constructs (i.e., trait anxiety, pain-related anxiety, and anxiety sensitivity) significantly and uniquely contributed to the prediction of negative affect at T1. Again, while the association among negative affect and trait anxiety (Holroyd et al., 1999), pain-related anxiety (Asmundson and Larsen, 2000), and anxiety sensitivity (e.g., Asmundson and Norton, 1995; Asmundson et al., 1999b) is supported by previous research, the present research suggests relationships are unique. The only anxiety construct to contribute significantly and uniquely to the prediction of perceived control was trait anxiety. Hallberg and Carlsson (1998) have also previously found an association between trait anxiety and perceived control. It is noteworthy that pain severity made unique, significant contributions to the prediction of each of the measures of functioning, explaining 33% of the variance in disability, 19% of the variance in negative affect, and 6% of the variance in perceived control. Anxiety, after controlling for pain severity, explained a much lower but, nevertheless, significant proportion of the variance in disability and negative affect Predicting functioning at three month follow-up The anxiety constructs found to predict T1 measures of functioning were not always the same variables found to predict functioning at 3-month follow-up (i.e., T2). With regard to disability, it was found that, after controlling for T1 pain severity and disability, only health anxiety significantly and uniquely predicted T2 disability. Patients found to exhibit higher levels of health anxiety tended to have a more difficult time performing work, marital, and social activities at 3-month followup. Interestingly, it was not health anxiety, but painrelated anxiety, that best predicted concurrent disability (i.e., at T1). This indicates that in the short-term, a measure focussing specifically on pain-related anxiety may contribute to our understanding of current functioning. However, in the long-term it may be a general tendency to worry about health that is predictive of long-term disability. Over time, health anxiety rather than pain-related anxiety may lead an individual to avoid specific activities that might otherwise have aided the healing process, resulting instead in increased disability. Health anxiety and anxiety sensitivity were found to be significant predictors of emotional distress (e.g., depressed mood, irritability, tension) at T2, after controlling for T1 pain severity and negative affect. Once again, it should be noted that the predictors of negative affect in the long-term differed from those found to be important in the short-term. Health anxiety only became important at T2. Anxiety sensitivity was the only construct remaining a significant predictor from short-term to long-term (perhaps not so surprising given its conceptualization as a relatively stable trait). The failure of trait anxiety to remain a significant predictor of negative affect is interesting. One potential explanation for its failure to explain a significant portion of the variance over the long-term is its lack of specificity to the pain experience. The same explanation could not be used for the failure of pain-related anxiety to significantly predict emotional distress in the long-term. Instead, it may be that pain-related anxiety is a less stable trait, for some, waning as time goes on or as the pain experience stabilizes. The significance of health anxiety, as previously stated, may also relate to continuous worry that oneõs health will worsen, resulting, in turn, in increased negative affect (Hadjistavropoulos et al., 2000a). Therefore, it may be that health anxiety serves to maintain pain and disability and that this, in turn, results in emotional distress. Anxiety sensitivity may similarly contribute to this cycle and, hence, the same outcome (i.e., negative affect) through a hypothesized association with increased fear of pain and associated avoidance (Asmundson, 1999). Problematic is the failure of anxiety sensitivity to also predict disability in the long-term. However, the association between anxiety sensitivity and pain may have less to do with physical disability than it does with negative thoughts and emotions. Plehn et al. (1998) findings lend support to this supposition as they did not find that the ASI significantly contributed to the variance in physical functioning, but that it explained the greatest portion of the variance in psycho-

9 H.D. Hadjistavropoulos et al. / European Journal of Pain 8 (2004) logical functioning, as measured by the SF-36 (McHorney et al., 1993). Further, the finding that anxiety sensitivity remained a significant predictor of negative affect supports other previous research indicating this same association, and lends support to the finding that the experience of pain is influenced by anxiety sensitivity and not vice versa (Keogh and Birkby, 1999; Keogh and Mansoor, 2001). Further research is needed to clarify these findings. This leads also to the conclusion that there is a need to examine more closely the interrelationships among the various anxiety constructs if we are to gain a better understanding of why these constructs are more-or-less important to the different measures of functioning and adaptation Summary and future directions After controlling for initial levels of functioning and pain severity, the addition of the anxiety constructs to the models predicting longitudinal functioning served to explain an additional 7% of the variance in disability and an additional 9% of the variance in negative affect. The amount of variance explained by anxiety in this study may ostensibly seem quite low, but considering that we used a conservative method of exploring variance in outcome attributable to anxiety (controlling for both level of functioning and pain severity), we believe this to be clinically significant. The amount of variance attributable to anxiety may also be lower than would have been found had we used a different measure of pain severity that did not share as much method variance with our measures of disability, negative affect, and perceived control. The results demonstrate that painrelated anxiety, health anxiety, trait anxiety, and anxiety sensitivity are interrelated but, importantly, somewhat unique in predicting adjustment to pain. Over an extended period, health anxiety appears to be the best unique predictor of disability, whereas both health anxiety and anxiety sensitivity are good predictors of negative affect. This does not imply, however, that the other anxiety constructs are not important to understanding pain and its consequences, as there may exist other reasons (e.g., treatment goals) for their measurement. Further, causality cannot be assumed from these results, as it may be that some other unaccounted variable is operating in this regard. One limitation of the present study is the 50% response rate at T2. Significant differences between the initial sample and those choosing to respond after 3-months were found with respect to age and education, but not among any of the main variables under investigation. In future research efforts, special attention should be given to ensuring that participants with lower levels of education and from younger age groups choose to remain in the study over the long-term so that any unique aspects of their pain experience can be accounted for. Another limitation is that all of the data collected was in self-report format and, therefore, raises the issue of method invariance. The addition of measures other than those based on self-report (e.g., behavioural indicators of pain-related anxiety) would serve to validate the current data. Finally, it should be noted that the 3-month period was selected arbitrarily. The ability of anxiety to predict adjustment over a longer period of time needs to be explored. Given the strong correlations among the anxiety constructs studied, and the fluctuating amounts of shared variance in predicting outcome variables, understanding the associations among these variables will be imperative to the advancement of knowledge in this area. To date, there has been some research into the hierarchical structure of fear and what are believed to be the fundamental fears (e.g., Taylor, 1995). But, researchers may want to include some of the other anxiety constructs studied herein as they have not previously been examined within such a hierarchical model. The present research suggests that both higher- and lowerorder anxiety factors explain variance in adjustment to pain. There is also a need to measure other constructs that might allow for the more accurate prediction of future disability, negative affect, and perceived control, as for each of these constructs there remained a great deal of unexplained variance. The nature of treatment is a likely and obvious determinant of outcome, but also other variables deserve further attention. For instance, other research with chronic pain patients has shown that extraversion and meaning variables differentiate treatment responders from poor responders (Kreitler et al., 1989), and that perceived competence influences various pain outcome measures (Moosbrugger and Schermelleh-Engel, 1991; Schermelleh-Engel et al., 1997). Interest in measures that assess individual differences in attention to pain (McCracken, 1997; McWilliams and Asmundson, 2001), as well as those that utilize other operationalzations of disability (e.g., work interference, social interference, days off work), may prove to be fruitful as well. Finally, in accord with biopsychosocial models of health and illness, biological and social factors also warrant careful consideration in this context (e.g., Engel, 1997, 1980). Application of these approaches, combined with the present findings, will extend our understanding of individual differences in response to the suffering associated with persistent pain and, importantly, will serve to refine our assessment and intervention strategies. At present, the results of this study serve to provide support for current clinical practice in which psychologists are frequently requested to address patient anxiety, including health anxiety, trait anxiety, pain-related anxiety, in the hopes of improving clinical outcomes (Hadjistavropoulos and Williams, in press).

10 10 H.D. Hadjistavropoulos et al. / European Journal of Pain 8 (2004) 1 11 Acknowledgements This research was funded by Health Services Utilization and Research Commission Establishment grant awarded to the first author. References Asmundson GJG. Anxiety sensitivity and chronic pain: empirical findings, clinical implications, and future directions. In: Taylor S, editor. Anxiety sensitivity: theory, research, and treatment of the fear of anxiety. Mahwah, NJ: Lawrence Erlbaum and Associates; Asmundson GJG. Commentary: anxiety sensitivity and the pain experience. Eur J Pain 2001;5:12 5. Asmundson GJG, Bonin MF, Frombach IK, Norton GR. Evidence of a disposition toward fearfulness and vulnerability to posttraumatic stress in dysfunctional pain patients. Behav Res Ther 2000;38: Asmundson GJG, Frombach IK, Hadjistavropoulos HD. Anxiety sensitivity: assessing factor structure and relationship to multidimensional aspects of pain in injured workers. J Occup Rehabil 1998;8: Asmundson GJG, Kuperos JL, Norton GR. Do patients with chronic pain selectively attend to pain-related information. Preliminary evidence for the mediating role of fear. Pain 1997;72: Asmundson GJG, Larsen DK. Pain anxiety symptoms scale. In: Maltby J, Lewis CA, Hill A, editors. Commissioned reviews on 300 psychological tests. Wales, UK: Edwin Mellen Press; Asmundson GJG, Norton GR. Anxiety sensitivity in patients with physically unexplained chronic back pain: a preliminary report. Behav Res Ther 1995;33: Asmundson GJG, Norton GR, Allerdings MD. Fear and avoidance in dysfunctional chronic back pain patients. Pain 1997;69: Asmundson GJG, Norton PJ, Norton GR. Beyond pain: the role of fear and avoidance in chronicity. Clin Psychol Rev 1999a;19: Asmundson GJG, Norton PJ, Veloso F. Anxiety sensitivity and fear of pain in patients with recurring headaches. Behav Res Ther 1999b;37: Asmundson GJG, Taylor S. Role of anxiety sensitivity in pain-related fear and avoidance. J Behav Med 1996;19: Burns JW, Mullen JT, Higdon LJ, Wei JM, Lansky D. Validity of the pain anxiety symptoms scale: prediction of physical capacity variables. Pain 2000;84: Burton AK, Tillotson KM, Main CJ, Hollis S. Psychosocial predictors of outcome in acute and subchronic low back trouble. Spine 1995;20: Costa Jr PT, McCrae RR. Neuroticism, somatic complaints, and disease: is the bark worse than the bite? J Pers 1987;55: Crombez G, Vlaeyen JWS, Heuts PHTG, Lysens R. Fear of pain is more disabling than pain itself: evidence on the role of pain-related fear in chronic back pain disability. Pain 1999;80: Engel GL. The need for a new medical model: a challenge for biomedicine. Science 1997;196: Engel GL. The clinical application of the biopsychosocial model. Am J Psychiatry 1980;137: Hadjistavropoulos HD. Cognitive and behavioural responses to illness information in health anxiety. Unpublished doctoral dissertation, University of British Columbia, Vancouver, British Columbia, Canada, Hadjistavropoulos HD, Craig KD, Hadjistavropoulos T. Cognitive and behavioural responses to illness information: the role of health anxiety. Behav Res Ther 1998;36: Hadjistavropoulos HD, Frombach IK, Asmundson GJG. Exploratory and confirmatory factor analytic investigations of the Illness Attitudes Scale in a nonclinical sample. Behav Res Ther 1999;37: Hadjistavropoulos HD, Hadjistavropoulos T, Quine A. Health anxiety moderates the effects of distraction versus attention to pain. Behav Res Ther 2000b;38: Hadjistavropoulos HD, Owens KMB, Hadjistavropoulos T, Asmundson GJG. Hypochondriasis and health anxiety among pain patients. In: Asmundson GJG, Taylor S, Cox BJ, editors. Health anxiety: clinical research perspectives on hypochondriasis and related disorders. London, UK: Wiley; 2000a. Hadjistavropoulos HD, Williams A. Psychological Interventions and Chronic Pain, In Hadjistavropoulos T Craig KD, editors. Pain psychological treatment perspectives. New Jersey: Lawrence Erlbaum, in press. Hallberg LR, Carlsson SG. Anxiety and coping in patients with chronic work-related muscle pain and patients with fibromyalgia. Eur J Pain 1998;2: Hitchcock PB, Mathews A. Interpretation of bodily symptoms in hypochondriasis. Behav Res Ther 1992;30: Holroyd KA, Malinoski P, Davis MK, Lipchik GL. The three dimensions of headache impact: pain, disability, and affective distress. Pain 1999;83: International Association for the Study of Pain. Classification of chronic pain. Pain 1986;(Suppl. 3):S1-S226. Kellner KA. Somatization and hypochondriasis. New York: Praeger- Greenwood; Kellner R, Abbott P, Winslow WW, Pathak D. Fears, beliefs, and attitudes in DSM-III hypochondriasis. J Nerv Ment Dis 1987;175:20 5. Keogh E, Birkby J. The effect of anxiety sensitivity and gender on the experience of pain. Cogn Emot 1999;13: Keogh E, Mansoor L. Investigating the effects of anxiety sensitivity and coping strategy on the perception of cold pressor pain in healthy women. Eur J Pain 2001;5: Kerns RD, Jacob MC. Assessment of psychosocial context of the experience of chronic pain. In: Turk DC, Melzack R, editors. Handbook of pain assessment. New York: Guilford Press; p Kerns RD, Turk DC, Rudy TE. The west haven-yale multidimensional pain inventory (WHYMPI). Pain 1985;23: Kreitler S, Carasso R, Kreitler H. Cognitive styles and personality traits as predictors of response to therapy in pain patients. Pers Indiv Differ 1989;10: Linton SJ, Skevington SM. Psychological factors. In: Crombie IK, Croft PR, Linton SJ, LeResche L, Von Korff M, editors. Epidemiology of pain. Seattle: IASP Press; p Linton SJ, Buer N, Vlaeyen JWS, Hellsing AL. Are fear avoidance beliefs related to the inception of an episode of back pain? A prospective study. Psychol Health 2000;14: McCracken LM. Attention to pain in persons with chronic pain: a behavioural approach. Behav Ther 1997;28: McCracken LM, Gross RT. The role of pain-related anxiety reduction in the outcome of multidisciplinary treatment for chronic low back pain: preliminary results. J Occup Rehabil 1998;8: McCracken LM, Gross RT, Aikens J, Carnrike Jr CLM. The assessment of anxiety and fear in persons with chronic pain: A comparison of instruments. Behav Res Ther 1996;34: McCracken LM, Zayfert C, Gross RT. The pain anxiety symptoms scale: development and validation of a scale to measure fear of pain. Pain 1992;50: McCracken LM, Zayfert C, Gross RT. The pain anxiety symptoms scale: a multimodal measure of pain specific anxiety symptoms. Behav Therap 1993;16: McHorney CA, Ware JE, Raczek AE. The MOS 36-item short-form health survey (SF-36): II. Psychometric and clinical tests of validity in measuring physical and mental health constructs. Med Care 1993;31: