Anxiety sensitivity, a fear of anxiety-related symptoms, has been associated with a heightened experience of pain, especially within women. The majority of experimental studies investigating this association have relied heavily on the cold pressor technique as a means of pain induction, limiting the generalisability of results.

The aim of the current study was to extend previous research by using two types of pain stimuli (cold and heat) to determine whether the link between anxiety sensitivity and pain generalises beyond cold pressor pain. The pain experience of 125 participants in response to these stimuli was assessed using threshold and tolerance readings, as well as subjective pain ratings.

Results indicated a positive association between anxiety sensitivity and subjective pain, with this association observed primarily in females. Although analysis also indicated a basic generalisability of results across pain stimuli, anxiety sensitivity effects appeared to be especially pronounced during heat stimulation.

These findings suggest that those high in anxiety sensitivity may respond more negatively to specific types of pain. Possible implications along with suggestions for future research are discussed.

Keywords: Anxiety sensitivity; Pain; Sex differences; Cold pressor; Thermal heat

Introduction

Anxiety sensitivity is defined as a trait tendency to be fearful of anxiety-related symptoms, based on the belief that such symptoms may indicate some impending harm (Reiss et al., 1986). Anxiety sensitivity is associated with a specific hypervigilance for bodily (somatic) sensations, along with an increase in the perceived magnitude of such sensations (Ball et al., 2002). In addition to links with emotional disorders such as panic, anxiety sensitivity has been closely associated with negative pain experiences in both chronic (e.g., [Asmundson and Norton, 1995] and [Plehn et al., 1998]) and acute settings (Lang et al., 2005). Similar findings have also been reported in laboratory studies in healthy volunteers. Keogh and colleagues, for example, found that anxiety sensitivity was positively associated with sensory and affective pain ratings in response to the cold pressor task ([Keogh and Mansoor, 2001] and [Keogh and Cochrane, 2002]). Interestingly, a study by Keogh and Birkby (1999) also found that anxiety sensitivity may have a sex-specific effect, with anxiety sensitivity differences in sensory ratings of cold pain found in females, but not in males.

In identifying a link between anxiety sensitivity and pain, the majority of these laboratory-based studies have tended to rely on a single method of noxious stimulation – the cold pressor task. However, as Janal et al. (1994) note, findings obtained using a specific pain modality have not always generalised to other pain modalities. Indeed, pain ratings compared across noxious stimuli such as cold pressor, heat, electrical and pressure pain responses have often demonstrated weak or even no correlation ([Davidson and McDougall, 1969], [Lynn and Perl, 1977] and [Janal et al., 1994]). Furthermore, Riley et al. (1998) found that while sex differences in pain tolerance were large for pressure pain, they were often small to moderate for noxious heat stimuli. More recently, Jones et al. (2003) found a positive association between trait anxiety and pain intensity during cold stimulation but no such association during heat stimulation. Such a comparison across pain types has not yet been made within the context of anxiety sensitivity, which is considered conceptually distinct from other negative emotional constructs, such as trait anxiety, fear of pain and pain-related anxiety ([Taylor, 1999] and [Keogh and Asmundson, 2004]).

The available data suggest that the investigation of anxiety sensitivity in relation to other types of pain induction is critical to determining the generalisability of effects. The primary aim of the present study was, therefore, to examine whether the relationship between anxiety sensitivity and pain was consistent across different types of pain stimuli. For the purposes of the current study, we chose to compare two related, yet different, types of pain (heat and cold) in healthy men and women who varied in levels of anxiety sensitivity. The specific predictions were that:

(1) Females would demonstrate greater pain sensitivity than males;

(2) Anxiety sensitivity and pain would be positively related, with this relationship expected to be particularly pronounced in females;

(3) The two relationships above would generalise across cold and heat modalities.

Method

2.1. Design
A regression design was employed. The two predictor variables were anxiety sensitivity and sex. The dependent variables were five measures of pain experience: two behavioural pain measures (pain threshold and tolerance time), and three subjective pain rating scales from the SF-MPQ (Melzack, 1987). Order of pain stimuli was counterbalanced within gender, with half of the participants receiving the cold followed by the heat stimulus, with this order reversed for the remaining half. Resting blood pressure, menstrual phase and negative mood state were also recorded, as previous research has suggested that these may operate as potential process variables underlying sex and anxiety sensitivity differences in pain ([Fillingim and Maixner, 1996], [Keogh and Birkby, 1999] and [Riley et al., 1999]).

2.2. Participants
The sample consisted of 125 adult participants (37 males and 88 females) with an overall average age of 23.9 (SD = 6.5). Participants were primarily Goldsmiths College undergraduate students recruited via the University’s course credits scheme, with the remainder paid volunteers (?5) who responded to advertisements posted around the university. Exclusion criteria were the presence of any condition that might have interfered with the detection of pain sensations (e.g., Raynaud’s disease, arthritis, cardiovascular disease), any current pain, or any prior medical history that may have resulted in a negative reaction to pain.

2.3. Pain induction
Two methods of pain induction were used in the current study: thermal cold and thermal heat. Thermal cold pain was induced using the cold-pressor task. This method was selected because of its high reliability and validity ([Wolff, 1984] and [Edens and Gil, 1995]) and to allow comparison with previous studies of anxiety sensitivity and cold pain ([Keogh and Birkby, 1999], [Keogh and Mansoor, 2001] and [Keogh and Cochrane, 2002]). To provide a common baseline, participants first placed their hand in a warm water tank (37 °C) for 2 min. Participants then transferred this hand to a cold-water tank maintained at a temperature of 1.5 °C (±0.2 °C), with a motor providing circulation of the water to prevent localised heating around the immersed hand. Participants reported when they first felt pain (‘threshold’), and when they were no longer able to tolerate the pain (‘tolerance’) at which point they removed their hand. An upper time limit of two minutes was imposed as around 8% of men and 4% of women appear to adapt to the increasing numbing effect of the cold without reporting pain (Wolff, 1984). Participants were unaware of the time limit prior to the pain task.

For the thermal heat induction, a Peltier-based thermode with a 5 cm ? 5 cm aluminum contact pad was used. After a visual demonstration by the experimenter, participants were required to undergo one practice trial and one experimental trial. For the practice trial, participants rested the thenar eminence of one hand (randomly selected) on the contact pad for 30 s, with the pad maintained at a constant temperature of 40 °C. This practice trial served to eliminate any substantial differences in baseline hand temperature across participants, and to familiarise participants with the heat sensation and thus reduce the likelihood of pre-tolerance instinctive withdrawal on the subsequent experimental trial. For the experimental trial, the pad was maintained at a constant temperature of 48 °C, with participants reporting threshold and tolerance in accordance with the protocol of the cold pressor task. An upper time limit of 10 s was imposed, with this limit established as appropriate during pilot testing. Participants were not made aware of this time limit prior to the task.

2.4. Questionnaires
A battery of measures was administered, with those relevant to the current study reported below:

2.4.1. Demographics
In addition to recording age and sex, the first page of the questionnaire battery included four items assessing compliance with restrictions (see Section 2.6). One participant reported having smoked in the hour prior to the experiment, and was re-tested later in the day. For females, menstrual phase data based on the estimated number of days since the start of the last/current period were also collected. Responses of 0–14 days were classified as ‘follicular’ with 15+ days classified as ‘luteal’ (Riley et al., 1999). Oral contraceptive use (pill/no pill) was also recorded.

2.4.2. Short form – McGill pain questionnaire (SF-MPQ; Melzack, 1987)
The SF-MPQ is designed to provide a comprehensive assessment of the sensory and affective dimensions of pain as well as an overall evaluation of pain intensity. The sensory pain scale consists of 11 sensory descriptors (e.g., ‘throbbing’), and the affective pain scale of 4 affective descriptors (e.g., ‘sickening’). Each descriptor is rated on a discrete 4-point scale from none (0) to severe (3). Overall sensory and affective pain scores can be calculated by aggregating scores on each subscale. A Present Pain Index (PPI) is also included which attempts to assess the overall level of pain intensity on a six-point scale ranging from 0 (‘no pain’) to 5 (‘excruciating pain’). The SF-MPQ correlates very highly with the full-length MPQ (Dudgeon et al., 1993) and is sensitive to change brought about by a variety of therapies ([Melzack, 1987], [Harden et al., 1991] and [Stelian et al., 1992]). Although sensory and affective subscales are frequently correlated, research has provided support for their demarcation as distinct factors (Wright et al., 2001).

2.4.3. Anxiety sensitivity index (ASI; Reiss et al., 1986)
The ASI is comprised of 16 items designed to assess the fear of anxiety-related sensations (e.g., ‘when I notice that my heart is beating rapidly, I worry that I might have a heart attack’). Respondents indicate the degree to which each item applies to them on a 5-point scale ranging from ‘very little’ (0) to ‘very much’ (4). Items on the ASI can be summated to give an overall anxiety sensitivity score or three separate subscales, often labelled ‘physical concerns’, ‘psychological concerns’ and ‘social concerns’ (e.g., McWilliams et al., 2000). Reliability data for the ASI are strong, with the scale displaying strong internal consistency (Ayvasik, 2000) and test–retest reliability ([Fullana et al., 2003] and [Lambert et al., 2004]). In addition, several studies have offered support for external validity of the ASI ([Maller and Reiss, 1987] and [Sandin et al., 1996]).

2.4.4. Depression anxiety stress scale-short form (DASS21; Lovibond and Lovibond, 1995b)
The DASS21 is an abbreviated 21-item version of the full length DASS (Lovibond and Lovibond, 1995a). Respondents are asked to indicate on a 4-point scale, the degree to which each item applied to them over the previous week (e.g., ‘I felt downhearted and blue’). Three separate subscale scores measuring the core symptoms of depression, anxiety and stress were calculated. A number of studies have supported this three factor structure ([Lovibond and Lovibond, 1995a] and [Crawford and Henry, 2003]), with the DASS displaying excellent reliability and validity in both clinical and community groups ([Antony et al., 1998] and [Clara et al., 2001]).

2.4.5. Coping strategy selection
At the conclusion of each pain task, participants were asked to indicate whether they had used a ‘focusing’, ‘distraction’ or ‘other’ coping strategy. This question was included as an exploratory item to investigate any sex or anxiety sensitivity differences in the type of coping strategy employed.

2.5. Blood pressure measurement
Resting blood pressure prior to pain testing was recorded with an Omron™ automatic blood pressure monitor with a self-inflating upper-arm cuff. Participants were seated and asked to relax for a period of around five minutes, in accordance with recommended guidelines (Shapiro et al., 1996), while the experimenter performed various administrative tasks. The cuff was attached to the participant’s non-dominant arm and three consecutive systolic blood pressure readings were taken, with an interval of approximately 2 min between each reading. An average blood pressure rating was calculated from these three readings.

2.6. Procedure
Following written consent, participants were asked to abstain from analgesics (48 h before), alcohol (12 h), caffeine (2 h) and nicotine (1 h) use prior to participation. The experimental sessions took place in a temperature-controlled room maintained at 22 °C. Three blood pressure readings were taken and demographic data recorded. Pain testing then commenced with participants exposed to a single pain stimulus applied to a randomly selected hand. Pain experience was assessed by recording threshold and tolerance times and by asking participants to complete the SF-MPQ immediately following noxious stimulation. The remaining questionnaire battery was then administered, and pain testing and assessment repeated on the other hand using the alternative pain stimulus. Finally, participants reported which coping strategy they had used during exposure to the pain stimuli. The duration of the whole experiment was approximately 35 min, by which time the effect of the pain manipulations had dissipated. The experimental procedures were approved by Goldsmiths College Ethical Committee and conformed to ethical guidelines for pain research with humans as recommended by the International Association for the Study of Pain (1995).

Results

3.1. Data screening
Cold pain tolerance times exhibited a bimodal distribution1. The second peak of the distribution represented 18 pain tolerance times of 120 s (the upper time limit), nearly three standard deviations above the mean. These outliers were trimmed to the next highest value (winsorising) as recommended by Tabachnick and Fidell (2001). Altering the value of the outliers in this way to improve analysis seems reasonable given that the value of 120 s is merely a product of a somewhat arbitrary cut-off time. This method of dealing with extreme case values also preserves the increased pain tolerance of the outlier sample, whilst ameliorating its disproportionate influence on the data. As bimodal pain tolerance data are commonly observed in studies of sex differences in pain tolerance, analysis using the winsorised distribution was compared to analysis using bootstrapped samples of the original bimodal distribution. Specifically, the p-value obtained from (1) a standard t-test (IV = sex) on the winsorised tolerance data was compared against the p-value obtained from (2) a t-test on the original bimodal data, but with p derived not from a standard t-distribution, but from a bootstrapped distribution of 10,000 resampled t-statistics of the bimodal data (e.g., Howell, 2002). Probability values were virtually identical across the two methods, supporting the validity of the winsorising method for the current data. The cold pain threshold data exhibited a similar, if far less pronounced, bimodal pattern and were trimmed in the same way.

After winsorising cold threshold and tolerance data, histograms revealed no univariate outliers. Mahalanobis distance based on three predictors was below the critical value of ?2(3) = 16.27, p

< .001 for every participant suggesting no multivariate outliers. A visual inspection of the histograms of all continuous variables showed that only the affective cold and heat pain measures demonstrated unacceptable departure from normality. As affective pain was positively skewed, a logarithmic transformation was applied to both scores (Tabachnick and Fidell, 2001). A check on the transformed variable showed an improved and acceptable approximation of normality. Scatterplots revealed no obvious non-linear relationships. Finally, residuals plots indicated that homoscedasticity assumptions were comfortably met for the regression analysis, and DeShon and Alexander’s (1996) test indicated that residual variances were equal across male and female groups, satisfying assumptions for the simple slope analysis.

3.2. Physiological controls
In order to determine whether female pain sensitivity was affected by hormonal factors, a series of 2 ? 2 ANOVAs with IVs menstrual phase (luteal/follicular) and pill type (no pill/pill) was conducted on all five cold and all five heat pain measures for the female sub-sample. Analysis revealed no significant differences on any pain measures, even with a liberal, uncorrected significance level of ? = .05. As these factors appeared to have no demonstrable impact on pain sensitivity, they were not considered in the main analyses.

Sex differences in resting blood pressure and mood were also investigated, in order to determine whether these factors could potentially underlie sex differences in pain sensitivity. An independent t-test on systolic blood pressure with sex as the IV revealed significant differences in blood pressure (t121 = 3.2, p < .005), with males (M = 120 mmHg) exhibiting higher resting blood pressure than females (M = 111 mmHg). t-tests revealed no sex differences on the DASS21 subscales. Correlations of anxiety sensitivity with both resting blood pressure and the DASS21 were also conducted and revealed significant associations of anxiety sensitivity with all three DASS21 scales (r = .39–.50, p < .01), although no association with blood pressure was observed. The potential role of both blood pressure and mood on pain rating will be considered in the main analysis below.

3.3. Analysis of order effects
We conducted mixed 2 * 2 ANOVAs on each of the five pain DVs2, with IVs of stimulus order (cold first vs. heat first; between-groups) and stimulation type (cold vs. heat; within-groups). However, no significant effects were observed, even at a liberal uncorrected significance level of ? = .05. This suggests no simple or differential carry-over effects were present (this is unsurprising given that the two trials were on different hands and with different stimuli). The addition of anxiety sensitivity3 and gender as IVs failed to reveal any further order effects.

We also conducted t-tests with stimulus order as the IV on the ASI and mood scales to assess whether questionnaire responses were influenced by stimulus type. No significant differences emerged.

3.4. Regression analysis on the five pain measures
A regression approach was used to investigate the relationship of anxiety sensitivity and sex with pain. Sex and anxiety sensitivity were centred (i.e. deviation scores computed by subtracting the mean of each variable from each individual score), with the interaction term the cross product of these centred variables. Centring was performed in order to reduce multicollinearity between the interaction and its constituent predictors, and to aid interpretation of the simple (main-effect) predictors (Aiken and West, 1991).

Separate standard regressions were performed on each of the five pain DVs with anxiety sensitivity, sex, and the anxiety sensitivity by sex interaction term entered as predictors. Separate regressions for each dependent pain measure were performed as, although naturally correlated, these measures nevertheless represent distinct pain dimensions (Price et al., 2001). To preserve the type I error rate, an ? = .015 was applied using the Dubey/Armitage–Parmar procedure (Sankoh et al., 1997). This procedure is based on Sidak’s formula, but with an adjustment for the average correlation between dependent variables4, and was carried out using Uitenbroek’s (1997) online SISA package. Results of the regression analyses are presented separately for cold and heat pain inductions.

Trevor Thompson, Edmund Keogh, Christopher C. French and Robert Davis

Department of Psychology, Goldsmiths College, University of London, New Cross, London SE14 6NW, UK
Department of Psychology, University of Bath, Claverton Down, Bath BA2 7AY, UK
Received 23 November 2006;  revised 15 February 2007;  accepted 13 April 2007.