gender moderates the association between depression and disability in chronic pain patients

www.EuropeanJournalPain.com

European Journal of Pain 10 (2006) 413–422

Gender moderates the association between depressionand disability in chronic pain patients

Edmund Keogh *, Lance M. McCracken, Christopher Eccleston

Pain Management Unit, Royal National Hospital for Rheumatic Diseases and University of Bath, UK

Received 10 January 2005; received in revised form 19 April 2005; accepted 29 May 2005Available online 11 July 2005

Abstract

Pain-related anxiety and depression are important correlates of disability amongst chronic pain patients. Furthermore, womenmay differ in their experience of pain, anxiety and depression when compared to men. The aim of the current study was to determinethe relative contribution of anxiety and depression on disability in male and female chronic pain patients. The sample consisted of260 patients (101 males, 159 females) referred to the Pain Management Unit at the Royal National Hospital for Rheumatic Diseasesin Bath, UK. As part of an initial assessment, all patients completed measures of depression, pain-related anxiety and disability. Aspredicted, both anxiety and depression were found to be significant positive predictors of pain, number of medications used anddisability. Although gender did not significantly predict disability, it did moderate the relationship between depression and disabil-ity, in that when depression was high, women report greater disability than men. Gender was also found to moderate the relation-ship between depression and number of medications used, in that a positive association was found for men, but not women.However, gender did not significantly moderate the relationship between anxiety and disability. Together these results not only sug-gest that gender is an important moderator of the relationship between emotional responses and disability, but that such associa-tions may be related more to depression than anxiety.� 2005 European Federation of Chapters of the International Association for the Study of Pain. Published by Elsevier Ltd. Allrights reserved.

Keywords: Pain; Gender differences; Anxiety; Depression; Disability

1. Introduction

Chronic pain is associated with more than just a neg-ative sensory experience (Price, 1999). Indeed, it is gen-erally accepted that negative emotional factors also playan important role in the perception and experience ofpain (Gatchel and Turk, 1999). Furthermore, there isgood evidence to suggest that alongside the pain itself,such negative emotional responses contribute towards

1090-3801/$32 � 2005 European Federation of Chapters of the International

reserved.

doi:10.1016/j.ejpain.2005.05.007

* Corresponding author. Present address: Department of Psychol-ogy, University of Bath, Claverton Down, Bath BA2 7AY, UK.Tel.:+44 1225 383671; fax: +44 1225 386752.

E-mail address: [email protected] (E. Keogh).

other aspects of chronicity, such as increased disability(Robinson and Riley, 1999).

There are numerous negative emotional states thatare associated with pain, with the main ones being an-ger, anxiety and depression (Robinson and Riley,1999; Blackburn-Munro and Blackburn-Munro, 2001,2003). Whilst there is only a limited amount of researchthat has considered anger (Greenwood et al., 2003),there is considerable research to suggest that anxietyand depression are associated with increased pain sensi-tivity and pain-related disability (e.g., Banks and Kerns,1996; Vlaeyen and Linton, 2000). It is not surprisingtherefore to discover that anxiety and depression areoften targeted in more psychologically based pain

Association for the Study of Pain. Published by Elsevier Ltd. All rights

mailto:[email protected]

414 E. Keogh et al. / European Journal of Pain 10 (2006) 413–422

interventions and that reduction in negative affect is of-ten related to improvements in symptoms (Morley et al.,1999). What is unclear is whether there are different rela-tions with the pain experience between anxiety anddepression when considering core influences on pain-related behaviours, since studies tend not to typicallytarget both mood states.

An additional robust and common finding is thatgender differences exist for pain and distress. Womenare generally found to report greater levels of pain, withgreater frequency and greater intensity when comparedwith men (Unruh, 1996; Berkley and Holdcroft, 1999;Fillingim, 2000). Furthermore, gender may also be animportant moderator of pharmacological and non-pharmacological pain interventions, as well as use ofhealthcare services (Weir et al., 1996; Edwards et al.,2000, 2003; Craft et al., 2004; Fillingim and Gear,2004). For example, we have recently investigatedwhether gender moderates aspects of outcome frominterdisciplinary pain management (Keogh et al.,2005). Although we generally found that there were nogender differences in pain outcomes such as disability,evidence was found to suggest that men and women dif-fered in current pain intensity, pain-related distress andpain catastrophizing (helplessness) at 3 months follow-ing treatment. At 3-month follow up, men maintainedpost-treatment reductions, whereas for women therewere no significant differences from pre-treatmentscores. Furthermore, such gender differences in painwere found to be mediated by distress andcatastrophizing.

Reasons for such differences are wide ranging and in-clude both biological and more psychosocial explana-tions. For example, the acquisition of gender rolesduring early stages of development have been used tohelp explain why there are gender differences in pain,as well as a range of other factors including mood andsocial functioning (Hankin and Abramson, 1999; Myerset al., 2003; Robinson et al., 2003). Gender role stereo-types are learnt early on in life, and can result in malesand females acting in gender-specific ways, and perhapseven result in treatment differences (e.g., Hibbard andPope, 1983; Piccinelli and Simon, 1997; Kroenke andSpitzer, 1998).

As mentioned above it seems that men and womenalso report differences in the experience of negativeemotional states, such as anxiety and depression (Kess-ler et al., 1994; Steiner et al., 2003; Weissman et al.,1996). The general pattern seems to be that on averagewomen are more vulnerable towards both anxiety anddepression, when compared to men. However, the effectsof experiencing anxiety and depression may be differentfor men and women, as a recent study reports thatwithin older adults, anxiety is associated with greatermortality in men but not women (Van Hout et al.,2004). Regarding the possible links with pain, such

observations have led to speculations that gender differ-ences in emotional vulnerability may be associated withgender differences in pain-related behaviours.

A few studies have directly examined gender differ-ences in pain and emotion (see, Bolton, 1994; Rileyet al., 2001; Edwards et al., 2000, 2003; McCrackenand Houle, 2000). In a study on non-specific back pain,Bolton (1994) found a relationship between depressionand disability within women, but not men. However,both Edwards et al. (2000) and McCracken and Houle(2000) report that the relationship between pain-relatedanxiety and pain severity was stronger in men thanwomen. In addition, Edwards et al. (2000) also exam-ined whether gender differences exist in the relationshipbetween depression and pain. Although depression wasindeed related to pain, this was similar in men andwomen. Finally, Riley et al. (2001) report a study usingstructural equation modelling in which they examinedgender differences in the relationship between used a ser-ies of visual analogue scales measuring pain (intensity,unpleasantness) and negative emotional responses topain (depression, anxiety, anger, frustration and fear).A stronger relationship was found between painunpleasantness and anxiety/depression for men whencompared to women.

Taken together these studies suggest that gender mayindeed moderate the relationship between pain and dis-tress. However, it is also apparent that the specifics ofthis relationship are unclear. For example, although itappears that men may be more vulnerable to the effectsof anxiety-related distress when in chronic pain, incon-sistencies are found when measuring depression. Thereare a number of reasons that might help explain thesediscrepant results. One issue relates to the assessmentof depression, which is often measured as a unitary con-struct using measures such as the Beck DepressionInventory (BDI; Beck et al., 1961). However, Morleyet al. (2002) have demonstrated through confirmatoryfactor analysis that within chronic pain patients, depres-sion, using the BDI, may be better conceptualised ascomprising of two factors: negative view of the self, so-matic and physical function. Also, some items weredropped due to their reduced reliability within chronicpain patients. It is therefore possible that issues associ-ated with measurement are causing some of theinconsistencies.

A second concern is that even in the few studies thatdo examine the relationship between gender and differ-ent types of distress (anxiety and depression) on pain,measures of pain-related behaviours have not beentaken. For example, although Riley et al. (2001) didexamine both anxiety and depression they only exam-ined the moderating effect of gender on the relationshipbetween emotion and pain; no measure of disability wasincluded. Given that some have found that pain-relatedbehaviours such as disability, rather than self-reported

E. Keogh et al. / European Journal of Pain 10 (2006) 413–422 415

pain, are clinically more important (e.g., McCrackenand Gross, 1998; McCracken et al., 2002) it would behelpful to determine whether such gender-related effectsexist for outcomes in addition to self-reported pain.

Finally, although it is assumed that gender moderatesthe relationship between distress and pain-related vari-ables, few studies directly adopt a moderation approachto determine whether the interaction between genderand emotion predicts pain-related outcomes (Baronand Kenny, 1986). Instead they rely on less direct meth-ods, such as differences in correlation coefficients. Theproblem with such indirect methods is similar to thoseinterested in interaction effects within analysis of vari-ance (ANOVA) models. Unless the interaction (or mod-eration effect) is directly examined, it is not possible toconfirm that one variable (e.g., gender) changes the pat-tern of the relationship between an independent (e.g.,mood) and dependent variable (e.g., pain). AlongsideANOVA, interaction effects can be examined using mul-tiple regression, which is useful when dealing with con-tinuous measures (such as questionnaire scores) as itmaintains statistical power (Maxwell and Delaney,1993).

In light of these shortcomings, the primary objectiveof the current investigation was to examine whether gen-der moderates the relationship between anxiety anddepression and pain-related variables. We hypothesisedthat:

1. anxiety and depression would be related to reportedpain and pain-related disability,

2. gender would moderate this relationship; given theresults from previous studies (e.g., Bolton, 1994;Edwards et al., 2000) it was hypothesized that anxietywould have a stronger positive association with painand disability in men than women, whereas depres-sion would be more strongly associated with painand disability in women than men.

2. Methods

2.1. Sample

All participants in the current study consisted ofchronic pain patients referred to the Royal NationalHospital for Rheumatic Diseases in Bath, UK. Of the260 patients, there were 101 males and 159 females,aged between 19 and 88 years (mean = 44.70;SD = 11.39). The majority of patients described them-selves as white (96%), and 13.5% reported either full-or part-time employment, with the majority (77%) notworking because of their pain. Patients also reportedbeing in pain between 12 and 600 months, with theaverage duration being 124.15 months (SD = 121.37months).

2.2. Procedure and measures

On referral, each patient was sent a battery of pre-assessment measures. The measures relevant to thecurrent study were:

The Beck Depression Inventory (BDI; Beck et al.,1961) was used as our primary depression measure.The BDI consists of 21-items, each comprising fourstatements (e.g., I am so sad or unhappy that I cannotstand it), with which patients indicate their level of expe-rience. Each item is scored on a 4-point scale (scored0–3). The BDI is one of the most commonly used mea-sured of depression, and possesses excellent reliabilityand validity. For the purposes of the current study wefollowed the recommendations of Morley et al. (2002)and constructed two subscales: negative self-evaluationand somatic/physical function. These subscales havebeen found to provide a much better representation ofdepression within chronic pain patients.

The Pain Anxiety Symptom Scale (PASS; McCrac-ken et al., 1992) was used to measure pain-related anxi-ety and avoidance. It consists of 40 items (e.g., I thinkthat if my pain gets too serve, it will never decrease),which participants are required to indicate their levelof experience when in pain. The items are scored on a6-point scale, with 0 indicating never, and 5 indicatingalways. The items can be used to calculate four sub-scales: fearful appraisal of pain, cognitive anxiety, phys-iological anxiety, and escape and avoidance behaviour.The PASS has good internal and external reliabilityand is considered a valid measure of pain-relatedanxiety.

Subjective pain was measured using an 11-pointnumerical scale, scored 0 (no pain) through to 10 (worstpossible pain). The question asked patients to indicatetheir current pain intensity, and was derived from thosefound on a number of standard pain measures, whichare considered valid and reliable methods of measuringpain intensity (Jensen et al., 1994, 1999).

Disability associated with pain was assessed using theSickness Impact Profile (SIP; Bergner et al., 1981). Thisscale consists of 136 behaviourally based items that mea-sure the impact of illness on 12 daily activities (e.g., Ispend much of the day lying down in order to rest). Pa-tients indicate whether or not each item applies to themthat day and are related to their health. The informationgained from this measure is combined to form a numberof sickness impact scales, which have been shown to beconsistent with other measures of disability. We focusedon the physical disability and psychosocial disabilitysubscales as these have been extensively used, especiallywithin chronic pain groups (e.g., Jensen et al., 1992).The SIP is believed to be a reliable measure with goodcontent, convergent and discriminant validity (de Bruinet al., 1992). Although some have questioned aspects ofthe measure (de Bruin et al., 1992; Pollard and John-


ston, 2001), the SIP is sensitive to the detection ofchange in clinical status in a range of different groups(Follick et al., 1985; Sullivan et al., 1990).

To ascertain the extent to which patients are engagingin other pain-related behaviours, each participant wasasked to indicate the number of hours per day spentresting/sleeping (rest) and number of hours slept at night(sleep). Patients also indicated their work status, numberof visits to general practitioners and number of differenttypes of medication used.

All procedures and measures outlined in this studywere reviewed and ratified by a local research ethicscommittee.

3. Results

Means and standard deviations for age, education,pain-related variables, anxiety, depression and disabilityfor men and women can be found in Table 1. Indepen-dent t-tests were conducted with gender (male vs.female) serving as the between-groups factor. None ofthe t-tests were significant, apart from one of the painanxiety scales (escape and avoidance) indicating thatgender differences were not generally evident on any ofthe variables.

It is possible that men and women differed with re-spect to the primary body location of their pain. Siteof primary pain was defined as either upper body (cer-vical, head, upper shoulder/limb), mid body (thoracic,lower back, abdominal) lower body (groin, lowerlimbs) or full body pain. Chi square analysis revealedthat there were no differences between men and womenwith respect to the location of primary body pain.

Table 1Mean and standard deviations (in parenthesis) for depression, anxiety, disab

Male

Age 43.33Years of education 11.82Number of pain-related surgeries 1.08Number of different Doctors seen for pain 5.63Number of visits to general practitioner in past 6 months 5.72Present pain rating 6.73Pain duration 110.45Medication use (no of different drugs) 2.41Hours resting per day 7.49Hours sleeping at night 4.96SIP physical disability 0.23SIP psychosocial disability 0.30BDI negative self-evaluation 5.14BDI physical and somatic 9.09PASS fearful appraisal 18.71PASS cognitive anxiety 29.09PASS escape and avoidance 25.93PASS physical anxiety 20.06

Note. BDI = Beck Depression Inventory; PASS = Pain Anxiety Symptom S* p < 0.05.

Analysis was also conducted to determine whetherthere were any differences in work status. Althoughthe majority of patients were not working due to pain,there were no gender differences with respect to thefrequency of patients that were working, not workingdue to pain or not working for other reasons (e.g.,retired).

3.1. Correlational analyses

One of the main predictions was that levels of anxietyand depression would be related to pain and pain-relatedvariables such as disability. In order to investigate this, aseries of correlations were conducted separately for menand women. As can be seen from Table 2 there were moresignificant correlations between the emotion scales andpain-related measures for women than men (e.g., visitsto general practitioners over the past 6 months, presentpain rating, hours resting per day). However, the stron-gest correlations found occurred for both men andwomen, and were between the emotion measures and dis-ability indexes. Pain chronicity was not significantlyrelated to any of the measures taken, apart from one(pain-related cognitive anxiety).

3.2. Moderating effects of gender

To determine whether gender moderates the relation-ship between anxiety and depression on the one hand,and pain and pain-related disability on the other, a ser-ies of moderated multiple regressions were conductedfollowing the procedures outlined by Baron and Kenny(1986). Alongside the two disability measures, we alsoexamined present pain and medication use since they

ility, pain, rest and sleep for men and women

Female t-Value

(11.99) 45.57 (10.94) �1.550(2.45) 12.31 (2.45) �1.560(2.43) 0.78 (1.47) 1.263(3.89) 6.03 (8.47) �0.443(6.55) 4.44 (4.43) 1.836(2.06) 6.48 (1.95) 0.973

(102.32) 132.85 (131.62) �1.454(1.35) 2.38 (1.19) 0.140(6.91) 6.01 (5.32) 1.861(1.95) 5.28 (2.21) �1.192(0.14) 0.24 (0.15) �0.428(0.16) 0.28 (0.17) 1.085(4.30) 4.97 (4.14) 0.310(3.37) 8.66 (3.41) 0.995

(10.60) 17.40 (10.06) 0.976(10.49) 28.50 (9.88) 0.451(9.06) 22.85 (10.49) 2.379*

(10.77) 19.30 (11.25) 0.524

cale; SIP = Sickness impact profile.

Table 2Correlations between pain-related measures and negative emotion scales for men and women

BDI: Self BDI: Som PASS: Fear PASS: Cog PASS: Escape PASS: Phy

Males

Number of pain-related surgeries �0.121 �0.254* �0.152 �0.136 �0.183 �0.001Number of Doctors seen for pain �0.036 �0.058 0.033 0.005 �0.204 �0.037Visits to general practitioner 0.262* 0.283** 0.129 0.178 0.084 0.271*

Present pain rating 0.105 0.252* 0.117 0.171 0.072 0.142Pain duration �0.104 �0.081 0.047 �0.027 0.032 �0.103Number of medications used 0.314** 0.257* 0.096 0.186 0.283** 0.206*

Hours resting per day 0.023 0.248* 0.177 0.113 0.199 �0.053Hours sleeping at night �0.157 �0.066 �0.018 0.073 0.276** �0.209*

SIP physical disability 0.165 0.364** 0.395** 0.242* 0.311** 0.305**

SIP psychosocial disability 0.358** 0.500** 0.525** 0.614** 0.319** 0.461**

Females

Number of pain-related surgeries �0.050 0.091 �0.005 �0.066 0.019 0.069Number of Doctors seen for pain �0.063 �0.037 �0.016 0.100 0.002 �0.007Visits to general practitioner 0.112 0.266** 0.232** 0.251** 0.175* 0.266**

Present pain rating 0.180* 0.350** 0.309** 0.227** 0.103 0.342**

Pain duration �0.064 �0.060 �0.120 �0.172* �0.130 �0.016Number of medications used 0.098 0.189* 0.232** 0.237** 0.290** 0.232**

Hours resting per day 0.217** 0.294** 0.215* 0.280** 0.287** 0.150Hours sleeping at night �0.233** �0.237** �0.132 0.031 �0.050 �0.145SIP physical disability 0.381** 0.503** 0.388** 0.254** 0.290** 0.416**

SIP psychosocial disability 0.678** 0.633** 0.531** 0.565** 0.328** 0.494**

Note. BDI = Beck Depression Inventory; Self = Negative self-perception; Som = somatic and physical complaints; PASS = Pain Anxiety SymptomScale; SIP = Sickness impact profile; Phy = Physical disability; Psych = Psychosocial disability; Hours = Hours resting during the day;Sleep = Hours sleeping at night.

* p < 0.05.** p < 0.001.


were the only other measures in Table 2 that correlatedabove 0.30 with pain-related anxiety and depression. Apreliminary investigation of the PASS revealed no differ-ences in the pattern of results for each subscale, and soan index of pain-related anxiety was calculated by sum-ming the four scales. Investigation of the BDI scales,however, revealed that interaction effects were foundfor the BDI – negative self-evaluation scale only. There-fore, this scale was used in the regression analyses.

Moderation was determined by calculating cross-product interaction terms between gender and the PASSindex, and between gender and BDI negative self-evaluation scale. Both product terms were calculatedfrom centred variables, which reduce problems associ-ated with multicollinearity (Aiken and West, 1991;Cohen et al., 2003). For each analysis hierarchical regres-sion was conducted entering the predictor variablesgender, centred PASS or centred BDI – negative self-evaluation scores at the first step, and the product termsof the two predictors at the second step. A significantchange in R2 between steps indicates a significant interac-tion (see Table 3).

The first analysis was conducted on current painscores. Both depression and anxiety were found to besignificant positive predictors of pain. However, theinclusion of the interaction terms did not significantlyimprove R2.

For number of medications used, both anxiety anddepression served as significant predictors. The interac-tion between depression and gender was close to anacceptable level of significance (p = 0.059), and so wasfollowed-up. Simple slopes were calculated and plotted(see Fig. 1) using the mean (±1SD) to create high, med-ium and low values for depression (Cohen et al., 2003).Following Holmbeck (2002) simple slope analysis re-vealed that the association between depression andnumber of medications used was positive and significantfor men (b = 0.099, t = 3.43, p < 0.001), but not for wo-men (b = 0.028, t = 1.19, p > 0.05).

For physical disability scores, both depression andpain-related anxiety were found to serve as significantpredictors. The interaction involving gender and depres-sion was significant, which was followed up with simpleslopes analysis (see Fig. 2). The association betweendepression and physical disability was positive and sig-nificant for women (b = 0.014, t = 5.27, p < 0.001), butnot for men (b = 0.005, t = 1.59, p > 0.05).

Finally, for psychosocial disability scores, both anxi-ety and depression were found to be significant predic-tors. Furthermore, the interaction term betweengender and depression accounted for a significantchange in R2. To examine this interaction further, simpleslopes analysis was conducted. The association betweendepression and psychosocial disability was positive and

Table 3Testing the moderating effect of gender on the relationship between distress (anxiety and depression) and pain-related variables (reported pain,medication use and disability)

Step Variables b t DR2 Total R2 F

DV = present pain

1 BDI:NS 0.149 2.408* 0.025* 0.025 3.225*

Gender �0.047 �0.7602 BDI:NS · Gender 0.057 0.579 0.001 0.026 2.256

1 PASS 0.234 3.714*** 0.056*** 0.056 7.169***

Gender �0.024 �0.3872 PASS · Gender 0.098 0.925 0.003 0.060 5.062**

DV = medication use

1 BDI:NS 0.190 3.077** 0.036** 0.036 4.744**

Gender �0.005 �0.0782 BDI:NS · Gender �0.184 �1.900a 0.014a 0.050 4.399**

1 PASS 0.259 4.135*** 0.067*** 0.067 8.611***

Gender 0.002 0.0322 PASS · Gender 0.036 0.344 0.000 0.067 5.759***

DV = SIP physical

1 BDI:NS 0.302 5.042*** 0.091*** 0.091 12.774***

Gender 0.028 0.4602 BDI:NS · Gender 0.198 2.113* 0.016* 0.107 10.121***

1 PASS 0.392 6.567*** 0.153*** 0.153 21.260***

Gender 0.057 0.9612 PASS · Gender 0.056 0.554 0.001 0.154 14.474***

DV = SIP psychosocial

1 BDI:NS 0.556 10.691*** 0.314*** 0.314 58.050***

Gender �0.059 �1.1342 BDI:NS · Gender 0.290 3.618*** 0.034*** 0.347 44.905***

1 PASS 0.583 11.056*** 0.339*** 0.339 61.591***

Gender 0.004 0.0682 PASS · Gender �0.022 �0.251 0.000 0.339 40.922***

Note. BDI:NS = Negative self-evaluation scale of the Beck Depression Inventory; PASS = Total score on the Pain Anxiety Symptom Scale;SIP = Sickness impact profile; BDI:NS · Gender = interaction term between BDI:NS and Gender; PASS · Gender = interaction term betweenPASS and gender.

a p < 0.06.* p < 0.05.

** p < 0.01.*** p < 0.001.

Fig. 1. Moderating effect of gender on the association betweendepression (negative self-evaluation) and number of different medica-tions used.

Fig. 2. Moderating effect of gender on the association betweendepression (negative self-evaluation) and physical disability.


Fig. 3. Moderating effect of gender on the association betweendepression (negative self-evaluation) and psychosocial disability.


significant for both men (b = 0.013, t = 4.14, p < 0.001)and for women (b = 0.028, t = 10.76, p < 0.001). Inspec-tion of Fig. 3 suggests that the positive relationship be-tween depression and psychosocial disability is greaterfor women than for men.

4. Discussion

We predicted that anxiety and depression would beassociated with a range of pain-related behavioursamongst a group of chronic pain patients. To thatend the evidence reported here certainly supports thisview. Specifically, we found that both pain-relatedanxiety and depression were related to pain, numberof visits to general practitioner, medication use,disability, rest taken during the day and sleep distur-bance. Such findings are entirely consistent with awide range of studies that have examined this typeof relationship (for review see Robinson and Riley,1999). It is now generally accepted that pain is morethan just a sensory experience, and that emotional fac-tors play a critical role in how pain is perceived, aswell as in how such chronic pain manifests itself interms of behaviour. This study confirms that bothpain-related anxiety and depression are strongly asso-ciated with pain-related behaviour.

Although the association between pain-related affectand pain behaviour was found, other differences com-monly reported in the gender and pain literature werenot found in this large chronic pain sample. In particu-lar, the predicted relationship between gender and painbehaviour was not found. Males and females did not dif-fer on any of the pain-related measures administered.This is out of line with previous data that report genderdifferences (e.g., Unruh, 1996; Berkley and Holdcroft,

1999; Fillingim, 2000). However, it should also beacknowledged that this finding is not alone in that somepublished studies also report not finding significantgender differences in the pain reports of chronic pain pa-tients (Robinson et al., 1998; Turk and Okifuji, 1999).Different also from previous studies was lack of any gen-der difference in pain-related anxiety and depression. Infact only one subscale of the anxiety measure differenti-ated between men and women, with women endorsing agreater likelihood of using escape and avoidance behav-iours in response to pain. Again previous research wouldcertainly lead us to expect that women would be foundto report higher levels of both anxiety and depressionthan men (e.g., Weissman and Olfson, 1995).

It remains unclear whether the lack of these previ-ously documented gender differences in pain popula-tions is a feature of this particular sample, or arisesfrom the measurement tools employed. First, the sampleitself is large and the analyses well powered. One featureof this chronic pain sample that is different from compa-rable samples is the extent of chronicity of complaint.Unusually, this is a complexly distressed and disabledpopulation referred to a national rehabilitation pro-vider. It is possible, and is deserving of further investiga-tion, that greater exposure to disability leads tocomparable levels of affective distress unaffected by gen-der. Second, it should be noted that other gender stud-ies, including our own, have used different affectmeasures such as the Anxiety Sensitivity Index (Keoghand Birkby, 1999; Keogh et al., 2004). It remains a pos-sibility for future research that some measures of affectare less affected by gender than others.

Main effects of gender aside, one of the main objec-tives of the current study was to investigate the potentialmoderating effect of gender on the relationship betweendistress and pain-related variables. Based on previousresearch (e.g., Bolton, 1994; Edwards et al., 2000) we ex-pected depression to be strongly related to pain vari-ables in women, whereas anxiety would be moreimportant in men. The findings partially supported thisprediction. Specifically, as expected the relationship be-tween depression and disability (both physical and psy-chosocial) was stronger for women than men. However,there was a slightly different pattern of effects for psy-chosocial and physical disability; depression was relatedto psychosocial disability in men and women (though toa greater degree in women). The association betweenphysical disability and depression only existed withinwomen, and not men, suggesting that depression mayhave a greater disabling effect for women. This resultis consistent with those reported by Bolton (1994). Suchfindings are not only consistent with the general viewthat women are more susceptible to depression, but alsosuggest that depression may leave women morevulnerable to disability. Such a link may help explainwhy women seem to suffer more pain-related problems.


We also expected pain-related anxiety to be morestrongly related to disability in men with chronic pain,when compared to women. Although pain-related anxi-ety (and depression) was related to pain and pain-relatedbehaviours, one noticeable difference was that genderdid not moderate any of the anxiety-pain relationships.This is inconsistent with previously published studiesthat have found that pain-related anxiety is morestrongly related to pain in men than women (Edwardset al., 2000; McCracken and Houle, 2000; Riley et al.,2001). This null finding is even more surprising giventhat we used a similar pain-related anxiety measure tothat previously used.

Reasons why such anxiety-related effects were notfound are unclear, especially given the relatively largesample size employed here. One observation is that theassociation between the anxiety and pain/disability-related variables for both men and women was muchsmaller in the current study when compared to others.It is therefore possible that the relatively weak relation-ship between the PASS and pain measures found heremay help to explain inconsistencies with previous find-ings relating to gender. However, this does not explainwhy there should be such weak relationships. One possi-bility could be due to problems with the scales used.However, this is unlikely given that all are well usedwithin pain groups, and such measures have been shownto possess good psychometric properties. Alternatively,differences between samples may be important. Muchof the previous research that has shown gender differ-ences in the association between anxiety and pain havebeen conducted on patients from North America,whereas the current study was based in the UK. Giventhat there are important differences between UK andUS healthcare systems, such differences may alter thepattern of results found. Such explanations are specula-tive and until further investigations are conducted, per-haps all that can be said at present is that genderdifferences may be particularly important when it comesto depression-related problems, but that for pain-relatedanxiety evidence is inconsistent.

The fact that we did not find evidence that gendermoderates the relationship between distress and currentpain was also interesting, especially given that womentend to report higher levels of both when compared tomen. This also suggests that the link between genderand pain reports may not always be reliable. That said,the fact that we found that gender moderates the linkbetween distress and disability, rather than current pain,may actually be of greater clinical significance (McCrac-ken and Gross, 1998; McCracken et al., 2002). However,given that measures of anxiety and depression have notbeen systematically examined in the context of genderdifferences in pain means that more research is required.

An additional finding was that with respect to thenumber of different medications reported being used

for pain, again the role of depression was different formen and women. Here we found that depression waspositively related to increased medication use in men,but not in women. This presents us with an intriguingsituation in which the interaction effect between genderand depression on pain-related variables depends onthe outcome being measured. Depression is morestrongly related to disability in women, yet seems tobe more strongly related to number of medications ta-ken in men. Reasons why such different patterns of re-sults should be found are unclear at present, althoughthey do highlight the critical need to clearly specifywhich outcome variable (e.g., self-reported pain, disabil-ity, medication use) is being considered. More generally,this finding also highlights that one should add apsychosocial (emotional, behavioural) component tostudies that seeks to investigate gender differences inanalgesic effectiveness (see, Fillingim and Gear, 2004).

Social gender roles may also provide some useful in-sight into the reasons why there are differences betweenmales and females in the relationship between mood andpain-related disability. For example, it could be that forfemales there is greater belief that a link exists betweendepressed symptoms and pain, which may not be as ac-cepted by men. Such gender-roles may also impact onhow healthcare practitioners treat men and women, witha greater willingness to prescribe anti-depression medi-cations to women, thus reinforcing the link betweendepression and pain for women. Further research is,therefore, required not only to confirm the consistencyof relationships found here, but also to consider the po-tential reasons why they may exist as well. The use ofmediation analysis (e.g., Keogh et al., 2004) would helpto confirm whether gender roles help explain the patternof results found.

Unfortunately, what the current study cannot tell usis whether the depression-related association with medi-cation usage in men is a product of physician prescribingbehaviours or due to patient behaviour. Furthermore,our reliance on self-reports of number of different med-ications used could be considered limited, and possiblyprone to a reporting bias. We only used a relativelysimple method of ascertaining medication use, which in-volved patients listing medications taken, and so lackedsophistication. Even so, the general notion that theremay be gender differences in medication-taking behav-iours is not considered controversial, and the results pre-sented here should be a good starting point for futureresearch.

A related criticism of this study is that it relied onself-report measures of anxiety and depression, whichtend to be highly correlated. However, current thinkingis that although they share a general tendency towardsdistress and negative thinking, anxiety and depressioncan be discriminated by considering the specific compo-nents of each (Clark and Watson, 1991; Mineka et al.,


1998; Keogh and Reidy, 2000). Similarly, it is possiblethat there is overlap between the disability and depres-sion measures. For example, the psychosocial disabilityscale may be viewed by some as just another measure ofnegative affect. If so, then it comes as no surprise to dis-cover that this component of disability is highly relatedto measures of depression. However, even if one is cau-tious about the findings relating to psychosocial disabil-ity, this does not detract from the fact that gender alsomoderated the relationship between depression and thephysical disability scale, where such an argument is lessvalid. Nevertheless future research may wish to developmethods to tease apart the specific components of anxi-ety and depression within pain groups, as well as con-sider the different aspects of disability and theirrelationship to mood.

The cross-sectional design of this study could also beproblematic in that it limits any consideration of causalrelationships. While we have conceptualized these con-structs in such a way as to argue that depression pre-dicts disability, it is possible that this relationship isbidirectional. The relative role that anxiety and depres-sion have on pain reports and disability over time is cur-rently unknown. Future research should consider longerterm, prospective investigations as the potential linksbetween distress and pain-related behaviours, andwhether men and women differ over the course of amuch longer time span. Further, the robustness of thesefindings will only be supported with replication in a dif-ferent setting.

If the current results are reliable, then there may be anumber of potential implications for tailoring interven-tions. The first is that if women really are more suscep-tible to greater disability when depressed, then theobvious target would be to consider focusing on depres-sion-like tendencies, and associated behaviours (e.g.,withdrawal, avoidance) when dealing with women. Sim-ilarly, it may be important to focus on depression whenconsidering medication use in men. An additional focusfor future research would be to determine what thepotential mechanisms that underpin such differencesbetween men and women regarding pain-related disabil-ity and depression. Although at present speculative, ithas been argued that hormonal factors may underpinsome of the differences observed between men andwomen with respect to the frequency and impact of bothpain and depression (e.g., LeResche et al., 2003; Black-burn-Munro and Blackburn-Munro, 2001, 2003; Steineret al., 2003). However, social and learning factors arelikely to also play a role in the association betweenpain-related disability and depression, and so shouldalso be the focus for future research. Prospective studieswould to help answer these interesting, yet speculative,questions.

In sum, the current study found evidence to suggestthat distress is related to a range of pain-related vari-

ables, including current pain and disability. We alsofound evidence to suggest that gender plays a role inmoderating the relationship between depression and dis-ability amongst chronic pain patients. Specifically, itseems that the relationship between depression and dis-ability is much stronger for women when compared tomen. This suggests that for women, targeting depressionmay help reduce disability in chronic pain patients.

References

Aiken LS, West SG. Multiple regression: Testing and interpretinginteractions. Newbury Park, CA: Sage Publications; 1991.

Banks SM, Kerns RD. Explaining high rates of depression in chronicpain: A diathesis-stress framework. Psychol Bull 1996;119:95–110.

Baron RM, Kenny DA. The moderator-mediator variable distinctionin social psychological research: Conceptual, strategic, and statis-tical considerations. J Pers Soc Psychol 1986;51:1173–82.

Beck AT, Ward CH, Mendelson M, Mock J, Erbaugh J. An inventoryfor measuring depression. Arch Gen Psychiatry 1961;4:561–71.

Bergner M, Bobbitt RA, Carter WB, Gilson BS. The Sickness ImpactProfile: development and final revision of a health status measure.Med Care 1981;29:787–805.

Berkley KJ, Holdcroft A. Sex and gender differences in pain. In: WallP, Melzack R, editors. Textbook of pain. 4th ed. London: Chur-chill Livingstone; 1999. p. 951–65.

Blackburn-Munro G, Blackburn-Munro R. Chronic pain, chronicstress and depression: coincidence or consequence. J Neuroendo-crinol 2001;13:1009–23.

Blackburn-Munro G, Blackburn-Munro R. Pain in the brain: arehormones to blame? Trends Endocrinol Metab 2003;14:20–7.

Bolton JE. Psychological distress and disability in back pain patients:Evidence of sex differences. J Psychosom Res 1994;38:849–58.

de Bruin AF, de Witte LP, Stevens F, Diederiks JP. Sickness ImpactProfile: the state of the art of a generic functional status measure.Soc Sci Med 1992;35:1003–14.

Clark LA, Watson D. Tripartite model of anxiety and depression:psychometric evidence and taxonomic implications. J AbnormPsychol 1991;100:316–36.

Craft RM, Mogil JS, Aloisi AM. Sex differences in pain and analgesia:The role of gonadal hormones. Eur J Pain 2004;8:397–411.

Cohen J, Cohen P, Aiken SG, West LS. Applied multiple regression/correlation analysis for the behavioural sciences. 3rd ed. London:LEA; 2003.

Edwards R, Augustson EM, Fillingim R. Sex-specific effects of pain-related anxiety on adjustment to chronic pain. Clin J Pain2000;16:46–53.

Edwards R, Augustson EM, Fillingim R. Differential relationshipsbetween anxiety and treatment-associated pain reduction amongmale and female chronic pain patients. Clin J Pain 2003;19:208–16.

Fillingim RB, editor. Sex, gender & pain. Progress in pain research andmanagement, vol. 17. Seattle: IASP; 2000.

Fillingim RB, Gear RW. Sex differences in opioid analgesia: Clinicaland experimental findings. Eur J Pain 2004;8:413–25.

Follick MJ, Smith TW, Ahern DK. The sickness impact profile: a globalmeasure of disability in chronic low back pain. Pain 1985;21:67–76.

Gatchel RJ, Turk DC, editors. Psychosocial factors in pain: Clinicalperspectives. London: Guilford Press; 1999.

Greenwood KA, Thurston R, Rumble M, Waters SJ, Keefe FJ. Angerand persistent pain: Current status and future directions. Pain2003;103:1–5.

Hankin BL, Abramson LY. Development of gender differences indepression: description and possible explanations. Ann Med1999;31:372–9.


Hibbard JH, Pope CR. Gender roles, illness orientation and use ofmedical services. Soc Sci Med 1983;17:129–37.

Holmbeck GN. Post-hoc probing of significant moderational andmediational effects in studies of paediatric populations. J PediatrPsychol 2002;27:87–96.

Jensen MP, Turner JA, Romano JM. Chronic pain coping measures:Individual vs. composite scores. Pain 1992;51:273–80.

Jensen MP, Turner JA, Romano JM. What is the maximum number oflevels needed in pain intensity measurement. Pain 1994;58:387–92.

Jensen MP, Turner JA, Romano JM, Fisher LD. Comparativereliability and validity of chronic pain intensity measures. Pain1999;83:157–62.

Keogh E, Birkby J. The effect of anxiety sensitivity and gender on theexperience of pain. Cognition & Emotion 1999;13:813–29.

Keogh E, Reidy J. Exploring the factor structure of the Mood andAnxiety Symptom Questionnaire (MASQ). J Pers Assess2000;74:106–25.

Keogh E, Hamid R, Hamid S, Ellery D. Investigating the effect ofanxiety sensitivity, gender and negative interpretative bias on theperception of chest pain. Pain 2004;111:209–17.

Keogh E, McCracken L, Eccleston C. Do men and women differ intheir response to interdisciplinary chronic pain management?. Pain2005;114:37–46.

Kessler RC, McGonagle KA, Zhao S, Nelson CB, Hughes M,Eshleman S, et al. Lifetime and 12-month prevalence of DSM-III-R psychiatric disorders in the United States. Results from theNational Comorbidity Survey. Arch Gen Psychiatry 1994;51:8–19.

Kroenke K, Spitzer RL. Gender differences in the reporting of physicaland somatoform symptoms. Psychosom Med 1998;60:150–5.

LeResche L, Mancl L, Sherman JJ, Gandara B, Dworkin SF. Changesin temporomandibular pain and other symptoms across themenstrual cycle. Pain 2003;106:253–61.

Maxwell SE, Delaney HD. Bivariate median splits and spuriousstatistical significance. Psychol Bull 1993;113:181–90.

McCracken LM, Gross RT. The role of pain-related anxiety reductionin the outcome of multidisciplinary treatment for chronic low backpain: Preliminary results. J Occup Rehabil 1998;8:179–89.

McCracken LM, Houle T. Sex-specific and general roles of pain-related anxiety in adjustment to chronic pain: a reply to Edwards etal. Clin J Pain 2000;16:275–6.

McCracken LM, Zayfert C, Gross RT. The pain anxiety symptomscale: development and validation of a scale to measure fear ofpain. Pain 1992;50:67–73.

McCracken LM, Evon D, Karapas ET. Satisfaction with treatment forchronic pain in a specialty service: Preliminary prospective results.Eur J Pain 2002;6:387–93.

Mineka S, Watson D, Clark LA. Comorbidity of anxiety and unipolarmood disorders. Annu Rev Psychol 1998;49:377–412.

Morley S, Eccleston C, Williams A. Systematic review and meta-analysis of randomized controlled trials of cognitive behaviour

therapy and behaviour therapy for chronic pain in adults,excluding headache. Pain 1999;80:1–13.

Morley S, Williams A, Black S. A confirmatory factor analysis ofthe Beck Depression Inventory in chronic pain. Pain 2002;99:289–98.

Myers CD, Riley JL, Robinson ME. Psychosocial contributions to sex-correlated differences in pain. Clin J Pain 2003;19:225–32.

Piccinelli M, Simon G. Gender and cross-cultural differences insomatic symptoms associated with emotional distress. An interna-tional study in primary care. Psychol Med 1997;27:433–44.

Pollard B, Johnston M. Problems with the sickness impact profile: atheoretically based analysis and a proposal for a new method ofimplementation and scoring. Soc Sci Med 2001;52:921–34.

Price DD. Psychological mechanisms of pain and analgesia. Seat-tle: IASP Press; 1999.

Riley JL, Robinson ME, Wade JB, Myers CD, Price DD. Sexdifferences in negative emotional responses to chronic pain. J Pain2001;2:354–9.

Robinson ME, Riley JL. The role of emotion in pain. In: Gatchel RJ,Turk DC, editors. Psychosocial factors in pain. Edinburgh: Guil-ford Press; 1999. p. 74–88.

Robinson ME, Wise EA, Riley JL, Atchison JW. Sex differences inclinical pain: A multisample study. J Clin Psychol Med Settings1998;5:413–24.

Robinson ME, Gagnon CM, Riley JL, Price DD. Altering gender roleexpectations: effects on pain tolerance, pain threshold, and painratings. J Pain 2003;4:284–8.

Steiner M, Dunn E, Born L. Hormones and mood: From menarche tomenopause and beyond. J Affect Disord 2003;74:67–83.

Sullivan M, Ahlmen M, Bjelle A. Health status assessment inrheumatoid arthritis. I. Further work on the validity of thesickness impact profile. J Rheumatol 1990;17:439–47.

Turk DC, Okifuji A. Does sex make a difference in the prescription oftreatments and the adaptation to chronic pain by cancer and non-cancer patients?. Pain 1999;82:139–48.

Unruh AM. Gender variations in clinical pain experience. Pain1996;65:123–67.

Van Hout HP, Beekman AT, De Beurs E, Comijs H, Van Marwijk H,De Haan M, et al. Anxiety and the risk of death in older men andwomen. Br J Psychiatry 2004;185:399–404.

Vlaeyen JWS, Linton SJ. Fear-avoidance and its consequences in chronicmusculoskeletal pain: A state of the art. Pain 2000;85:317–32.

Weir R, Browne G, Tunks E, Gafni A, Roberts J. Gender differencesin psychosocial adjustment to chronic pain and expenditures forhealth care services used. Clin J Pain 1996;12:277–90.

Weissman MM, Olfson M. Depression in women: Implications forhealth care research. Science 1995;269:799–801.

Weissman MM, Bland RC, Canino GJ, Faravelli C, Greenwald S,Hwu HG, et al. Cross-national epidemiology of major depressionand bipolar disorder. JAMA 1996;276:293–9.

gender moderates the association between depression and disability in chronic pain patients

Documents