1.1. Pain catastrophizing
Pain catastrophizing is one of a number of concepts describing the psychological experience of pain. Catastrophizing was originally conceptualised as a cognitive error in the cognitive formulation of depression (Beck,1 [p.745]), but in pain is applied to appraisal of the meaning of pain, and is described by Sullivan et al.27 as “an exaggerated negative mental set brought to bear during actual or anticipated painful experience” (p.53).
1.2. Established psychometric properties of the Pain Catastrophizing Scale
The Pain Catastrophizing Scale (PCS) is widely used as the “reference standard psychometric tool for pain catastrophizing.”14 Respondents are asked to rate 13 pain-related statements on a 5-point Likert scale. During its development, Sullivan et al.25 investigated the factor structure in a sample of 439 students. Principal component analysis established that the PCS assessed 3 related dimensions of magnification (evaluation of the pain as a threat), rumination (repeated worry), and helplessness (belief that nothing can help to resolve the pain). Confirmatory factor analysis has since been used in English and Dutch versions to confirm this 3-factor structure compared with a unidimensional or 2-factor structure in students,19 community and pain outpatient samples,18 and pain-free students, patients with chronic low back pain, and patients with fibromyalgia.30 Overall, these studies demonstrate consistency of the 3-factor model of pain catastrophizing across participant groups in English and Dutch versions of the questionnaire.
Existing data on the reliability of the PCS report adequate to excellent internal validity scores (coefficient alphas: total PCS = 0.87-0.93, rumination = 0.87-0.91, magnification = 0.66-0.75, and helplessness = 0.78-0.8719,24). Pedler reported in a commentary review that “there are currently little data available regarding the test-retest reliability, sensitivity to change, and clinically meaningful change of the PCS” and that “[f]urther research investigating these dimensions of the PCS would significantly increase the clinical utility of this tool”20 (p.137). More broadly, self-report measures used in health care, including pain, tend to be developed using sample sizes too small adequately to establish relationships between the scale and population characteristics such as age, sex, and diagnostic category16 (p.34). Such thorough validation and reliability testing across wide samples of participants is possible using meta-analytic methods once the scale is widely used.
1.3. Pain catastrophizing scores and personal characteristics
Studies have been conducted to explore potential differences in pain catastrophizing between people of different ages, sexes, from different cultural backgrounds, and with different pain diagnoses. Such information helps to explain the extent to which the construct of pain catastrophizing can be viewed as stable across populations. Studies report either no sex difference23 or that women tend to score higher than men.5,13,25,28,29 Older adults tend to score lower than younger adults,12,26 but the effect is reversed in adolescents, where older adolescents score higher than younger.2 Although a number of studies have reported pain catastrophizing scores for participants with different pain diagnoses (including many studies in this review), no review or commentary has compared or combined these.
There have been no studies of the difference in pain catastrophizing scores of participants using different language versions of the PCS or other measures of pain catastrophizing. Any disparity in PCS scores of participants using different language versions could be due to translations of the outcome measure or to cultural differences in the experience and expression of pain catastrophizing. Existing studies report higher levels of pain catastrophizing in Chinese Canadians compared with European Canadians10 and in African Americans compared with white Americans5 using the same language scale. Therefore, some limited evidence from healthy participants suggests the presence of cultural factors in mediating pain catastrophizing scores.
On a subjectively weighted/majority basis, narrative reviews have concluded that women score on average higher than men. Otherwise, the rather inconsistent findings from studies on specific participant groups leave open the question of whether there are systematic age, sex, or other demographic differences in pain catastrophizing; no comprehensive investigation of such differences in pain catastrophizing has been attempted.
1.4. Aims of this review
The aims of this review were to systematically obtain data on PCS scores from a wide range of studies and first to explore the psychometric properties of the PCS using meta-analytic methods and then to investigate whether there were any clear differences in PCS scores according to personal characteristics of participants. The review also served to test the use of meta-analysis to investigate psychometric properties of self-report instruments widely used in psychological treatment.
2.1. Protocol and registration
The research protocol for the review and meta-analysis was registered on PROSPERO (prospective register of systematic reviews) at the University of York's Centre for Reviews and Dissemination, registration number CRD42016032863.
2.2. Eligibility criteria
English language studies reporting baseline PCS scores (those collected before any intervention) were included in the meta-analysis. Participants in those studies had to be aged 18 years or older and could have any health condition or none, and both randomized and nonrandomized studies were included. Included studies were those that stated in the title or abstract that they used the PCS and reported demographic and clinical information about participants (age, sex, and diagnostic category) and psychometric data for PCS scores (mean, SD, and sample size).
2.3. Search strategy
The search strategy was adapted for Cochrane Library, CINAHL, Embase, PsycInfo, PubMed, and Web of Science (all 1995-present) by using wildcards and terms relevant to each database (an example search strategy is included in supplemental Fig. 1, available at http://links.lww.com/PAIN/A735). The last search was run on 30 November 2015. Requests were sent to authors for data missing from otherwise relevant studies, resulting in 8 responses with eligible scores to 81 requests for missing PCS data and no eligible scores for 21 requests sent for missing demographic data.
2.4. Study selection and data collection
One reviewer (C.H.B.W.) screened the title and abstract of the studies retrieved in the database searches. A random sample (using a random number sequence generator) of 5% of the articles were screened by title and abstract by a second reviewer (S.J.M.), and the interrater reliability was calculated. Discrepancies were discussed and resolved. The data collection form was adjusted following the pilot to allow for pooled data from studies that reported only PCS subscores or scores from subgroups but personal data from the whole sample. Data items extracted from data including participant characteristics, study data, and study type are presented in supplemental Table 1 (available at http://links.lww.com/PAIN/A735).
2.5. Risk of bias in individual studies
A component approach was used to assess the risk of bias in each included study.15 Relevant components from the Quality Assessment Tool for Observational Cohort and Cross-Sectional Studies17 were used in this review and meta-analysis:
- (1) Was the study population clearly specified and defined?
- (2) Was the participation rate of eligible persons at least 50%?
- (3) Were all the subjects selected or recruited from the same or similar populations (including the same period)? Were inclusion and exclusion criteria for being in the study prespecified and applied uniformly to all participants?
Sensitivity analysis was conducted to ascertain the impact of bias from the above components on the overall effect sizes found in the meta-analysis. Meta-analysis of effect size was conducted first for all studies and then repeated only for studies known to be eligible (meeting all 3 risk of bias criteria), following the Cochrane method.8 There was unlikely to be a high risk of publication bias in the included data, so no risk of bias analysis across studies was undertaken.
2.6. Meta-analysis to explore the psychometric properties of the Pain Catastrophizing Scale
2.6.1. Calculating weighted mean values, SDs, and reliability alphas of Pain Catastrophizing Scale scores
Weighted scores were computed for the PCS mean, SD, and Cronbach alpha for each sample in which these data were available. Weights were based on the standard error for each sample. The weighted scores were used to compute the mean, SD, reliability coefficient, confidence intervals (CIs), and random-effects variance components for PCS scores across studies.
2.6.2. Reliability estimates
The internal consistency reliability for the PCS and its subscales was calculated by finding the weighted mean of the Cronbach alpha statistics reported in studies using the PCS. The test–retest reliability for the total PCS scale was calculated using the weighted mean test–retest reliabilities reported in studies.
2.6.3. Subgroup analysis
Wilson macros for SPSS31,32 were used to conduct Hedges–Olkin random-effects meta-analysis6 on participants grouped by pain diagnosis. Hedges and Olkin's method of meta-regression was chosen for this analysis because of its coherence with the theory of data used throughout: that including all available data and accounting for bias through weighting provides a more comprehensive analysis than excluding data. Hedges and Olkin's method uses a pooled variance estimate to standardize the difference between group means. Biases were corrected based on a sample size statistic using weighted scores.
A Q statistic was calculated to obtain a test of the homogeneity of the effect size (the extent to which individual effect sizes vary around the mean effect size); it is the standardized sum of squared differences between each effect size and the mean effect size:
where k is the number of studies or samples included, d+ is the mean effect size, and
is the weighted average based on the variance of the unbiased effect sizes.
2.6.4. Exploration of the heterogeneity of the mean Pain Catastrophizing Scale score across studies
The I2 measure of heterogeneity was calculated for the grand mean PCS score and for the mean PCS score of diagnostic subgroups. It was necessary to transform the Q value reported in the original meta-analysis to an I2 value owing to Q having “too much power as a test of heterogeneity if the number of studies is large”8 (9.5.2).
The I2 value was calculated from Q as follows:
I2 is reported as a percentage, where over 75% indicates substantial heterogeneity between trials.9
2.6.5. Use of multiple regression to explore heterogeneity of Pain Catastrophizing Scale scores
SPSS version 2311 was used to conduct random-effects meta-analysis and meta-regressions. Multivariate meta-regression was conducted to explore the heterogeneity of mean PCS scores across participant groups by testing their association with variables and other study features.
Wilson's macro for SPSS was used to employ Hedges and Olkin's psychometric meta-analysis method, with results transformed from Fisher z scores back to r scores after the analysis. This method of meta-regression uses a weighted least squares procedure and scores from each study that are weighted by the inverse of the study's sampling error bias. Variables entered into the first meta-regression were pain category (type of pain diagnosis), mean age of participants, proportion of female participants, year of study (studies were categorized into ranges of 3 years), study type, and language of PCS used. A further meta-regression was then run using pain category as the only variable. Rerunning the meta-regression with this clinically relevant variable meant that more studies were included, because some studies were excluded on the grounds of missing data—including those with no data on the sex of participants—in the first meta-regression.
3.1. Study selection
Two hundred twenty studies were identified for inclusion in the review and meta-analysis (see Fig. 1 for searching and screening process). Supplemental Table 2 provides details of reasons for the exclusion of studies (available at http://links.lww.com/PAIN/A735).
Interrater reliability was calculated for the screening of articles that was completed by the 2 independent raters (C.H.B.W. and S.J.M.). There was 90.3% agreement, with a Cohen kappa of 0.87 (accounting for agreement due to chance), meeting criteria for reliable interrater agreement4 (p.56). Discrepancies were discussed and resolved between the raters, with a conservative approach followed to allow for further screening at a later stage.
3.2. Data cleaning and preparation
Data cleaning was conducted to remove double-counted data and data with errors (2 articles in total). Seven further articles contained surprisingly low PCS scores but no evidence of the source of error; these articles were not removed from the analysis. Instead, meta-analytic methods were applied to correct for artifacts and error.
3.3. Study characteristics
Data from 220 studies published between 1997 and 2015 were included in the initial analyses. Included studies were cross-sectional, psychometric, case series, randomized controlled and nonrandomized controlled trials, case controlled, and cohort studies. Sample sizes ranged from 3 to 1786, and many studies reported PCS scores and related data for 2 or more groups of participants, so data were collected for 329 groups altogether. The PCS was represented by 21 languages translated from English.
Mean ages of participants in studies ranged from 19 to 76 years, with a grand mean age of 45 years, SD = 12 (for both weighted by sample size and unweighted). The grand total number of participants across included studies was 42,976. The sex ratio was 55:32:13 for men, women, and participants whose sex was not reported.
Mean PCS scores across all participant groups ranged from 3.2 to 43.8, with a grand weighted mean of 20.22 using a random-effects model (weighted SD = 10.26, 95% CIs of mean = 19.30-21.14). Unless otherwise stated, “PCS score” refers to the total scale score. Subscale scores are reported where available. Results of individual studies are presented in supplemental Table 1 (available at http://links.lww.com/PAIN/A735) because of the large number of studies (220) and larger number of participant groups in the studies (k = 339).
3.4. Risk of bias analysis
Three screening questions were used to assess the risk of bias within studies:
- Q1 = Was the study population clearly specified and defined?
- Q2 = Was the participation rate of eligible persons at least 50%?
- Q3 = Were all the subjects selected or recruited from the same or similar populations (including the same period)? Were inclusion and exclusion criteria for being in the study prespecified and applied uniformly to all participants?
Seventy studies fulfilled the criteria for all 3 screening questions. The weighted PCS scores for all studies included in the review and for just those studies meeting all the risk of bias criteria were calculated. Subgroup analysis determined the difference in PCS scores between studies that did and did not meet all risk of bias criteria; results are presented in Table 1.
Regression analysis showed the PCS score to be significantly related to whether a study met all risk of bias criteria, B = 16.64, SE = 0.49, 95% CI = 15.68 to 16.60, P < 0.001. Analysis of variance showed a significant correlation between the type of study conducted and whether the study met all risk of bias criteria, B = 5.62, SE = 0.12, 95% CI = 5.38 to 5.85, P < 0.05.
3.5. Meta-analysis to explore the psychometric properties of the PCS
3.5.1. Heterogeneity of the grand mean Pain Catastrophizing Scale score
The I2 value of the grand mean PCS score is 98.96%, representing very substantial heterogeneity between studies. The high I2 value might also suggest that the overall mean ES is misleading because there are subpopulations of studies represented that have different ES values; this supports the need to conduct subgroup analysis to further determine the origins of heterogeneity of mean PCS scores across participant groups (see section 3.5.4).
Estimates of the internal consistency of the PCS full scale and subscales were based on Cronbach coefficient alpha.3 After weighting and averaging reports from 40 studies, the full-scale alpha = 0.92 (95% CI = 0.91-0.93). PCS subscale data were reported in 21 studies, with average internal consistency of alpha = 0.89 (95% CI = 0.87-0.91) for the rumination subscale; alpha = 0.77 (95% CI = 0.73-0.82) for the magnification subscale; and alpha = 0.88 (95% CI = 0.86-0.9) for the helplessness subscale.
Six studies provided 8 samples (n = 317) that, when weighted and combined, produced a mean test–retest reliability of 0.88 (95% CI = 0.83-0.93, range 0.73-0.97), representing good reliability. The time lapse between the test and retest in included samples ranged from 7 to 135 days. It was not possible to test the reliability of scores by time lapse because a number of studies had a range of intervals between tests rather than a standardized interval (including 1 study in which the time lapse ranged between 14 and 135 days).
3.5.3. Analysis of heterogeneity of Pain Catastrophizing Scale scores across participant groups
Participants from studies included in this meta-analysis were categorized based on their pain diagnosis. The “other” group consisted of diagnoses that did not fit into one of the specified categories, including participants with mixed pain diagnoses; these were excluded from analysis of heterogeneity. The wide spread between branches in the plot of weighted mean PCS scores suggests substantial heterogeneity in the PCS scores across participant groups (Fig. 2). Notably, participants with lower limb pain experienced, on average, lower pain catastrophizing than healthy participants by 2 points of a possible score of 52 on the PCS, and the mean PCS scores of participants with upper limb or upper and lower limb pain were equivalent to those of healthy participants.
3.5.4. Subgroup analysis
Owing to the extent of heterogeneity between PCS scores of participants with different pain diagnoses, subgroup analysis was conducted to establish the heterogeneity of scores within diagnoses, to distinguish it from sampling error7 (Table 2). I2 values ranged from 92.27% to 99.04%, indicating high levels of heterogeneity within diagnostic groups; 193 groups (24,546 participants) fell into “other”/healthy/groups with mixed or unclear diagnoses and were excluded from this analysis.
Subgroup analysis between study types showed considerable overlap and homogeneity in mean PCS scores, with the exception of nonrandomized controlled trials, which had a higher mean PCS score than the grand total PCS score (27.55, 95% CI = 24.15-30.95 compared with 20.22, 95% CI = 19.44-21). This may be an artifact of data coming from 2 groups within 1 study, which recruited participants only if they “reported high levels of pain catastrophizing”22 (p.859).
3.5.5. Meta-regression of Pain Catastrophizing Scale scores
Multivariate meta-regression analysis was conducted to establish the association between PCS scores and characteristics of study participants (age, sex, and diagnostic category) and study design (language of PCS, type of study, and year of study publication). After exclusion because of missing data, 277 groups were included in the analysis. Diagnostic category of participants, language, and type of study were all significantly associated with the mean PCS score obtained (Table 3). In a further meta-regression conducted with the variable of diagnostic category, this variable was again significantly associated with the mean PCS score (all 329 participant groups were included in this analysis; Table 4).
Significantly higher or lower mean scores occurred for some non-English language versions of the PCS, notably for the Cantonese version of the PCS (the weighted mean score for Cantonese was 36.3 compared with a mean of 18.48 for English versions, β 15.31, P = 0.002). The Cantonese scores were based on 1 study using the Cantonese language version.
4.1. Psychometric properties of the Pain Catastrophizing Scale
The review included 220 studies of a total of 329 participant groups (42,976 clinical and nonclinical participants) with a mixture of pain diagnoses, age, and sex distributions, with a range of study types. Systematic review and meta-analytic methods were used to synthesize existing data by establishing and refining the known psychometric properties of the PCS using a far larger data set than previous studies. Internal consistency and test–retest reliability of the overall score and of the rumination subscale (but not of magnification or helplessness subscales) were all >0.85, encouraging confidence in use (see also Sullivan et al.25). The higher test–retest reliability found in this meta-analysis than previously25 could be explained by shorter intervals between testing or the inclusion in the current meta-analysis of stabler nonclinical samples alongside clinical samples. The psychometric findings on the PCS from a much larger evidence base may in part allay concerns over the “ad hoc” creation of psychometric self-report scales in health and psychology.16
4.2. The relationship between Pain Catastrophizing Scale scores and participant characteristics
Subgroup analysis showed that PCS scores did not differ systematically with age or sex, again encouraging confidence in use, particularly in the light of contradictory results of previous studies.2,5,12,13,23,25,26,28,29 However, different pain diagnoses were associated with differences in PCS scores, and although this could be mediated by the association of pain catastrophizing scores with pain intensity reports,21 it is more likely that different pain problems are associated with different specific worries, and those with diagnoses that are difficult to establish or contested (such as fibromyalgia33) may more consistently generate catastrophic thinking. This requires further investigation.
The findings of this review reveal differences in PCS scores across populations and languages, but given the complexity of cultural and linguistic translation, it would be premature to try to interpret these findings. They do, however, serve as a reminder that simply translating a questionnaire does not guarantee that its psychometric properties remain the same as in the original language.
4.3. Strengths and limitations of the review
Strengths and limitations of this review are considered within a broader synthesis of the use of meta-analysis to establish psychometric properties of a self-report scale. The use of meta-analysis provides a transparent process that can be replicated or updated by anyone at any time, potentially providing a continually updated set of psychometric data. Meta-analysis is intended to give an unbiased summary based on a larger sample than narrative reviews.
Studies were included in this review if PCS use was reported in the study abstract, so we are reasonably confident that all large-scale psychometric studies were included. Studies that used the PCS but did not state that they did so in the title or abstract may have been missed by this criterion. Maximum use was made of included studies by including all data available per analysis. In addition, the principle of assessing rather than selecting against methodological deficiency allowed the inclusion of all studies regardless of risk of bias, examined instead by sensitivity analysis and by weighting scores using a random-effects model in the meta-analysis.
In this review, participants were categorized according to a single pain diagnosis, but many people with chronic pain have more than 1 pain condition and could not be fitted into a pain diagnosis category. Studies used different ways of categorizing pain or describing pain diagnoses, meaning that data were matched to “best fit” for this meta-analysis, for example, “low back pain,” “acute low back pain,” “chronic low back pain,” and “persistent nonspecific low back pain” were classified as “lumbar pain,” although there may have been differences in diagnosis and threshold for classification in the different studies. The high proportion of “healthy” participants in included studies who were students limits the generalizability of the results to the general population of “healthy” people without a pain diagnosis but with a wider age range. This is not unique to this review.
It is possible that the decision to include other language versions of the PCS introduced biases and inaccuracies owing to different psychometric properties of these versions. The decision was justified by the widespread international use of the PCS, highlighting the need for further validation studies of the translated versions of the questionnaire.
The risk of bias screening was completed by 1 author (C.H.B.W.). Optimally, a second author would duplicate the screening, and results would be compared. Otherwise, established protocols for systematic review and meta-analysis were followed.15 Meta-analytic methods were used to correct for measurement artifacts within included studies by weighting scores to obtain more accurate estimated effect sizes.
Samples used in the regression analysis were not fully independent in that, frequently, more than 1 participant group was included from a single study. This increased the number of groups available to analyze, but the results should be treated with caution because of this nonindependence of samples.
Overall, this review demonstrates a comprehensive attempt to identify relevant articles and a systematic method of discussing and deciding on inclusion and exclusion of studies. The use of meta-analysis across a large sample of studies allowed for exploration of narrow variance of PCS scores. The findings provide greater support for evidence from previous individual studies on the reliability and validity of the PCS.
4.4. Implications for clinical practice and research
Studies in this meta-analysis highlighted that the PCS is widely used for research and clinical practice. Current normative values and clinical cutoff scores are based on a sample of 851 injured workers, 75% of whom had a soft tissue back injury24 (p.6). This meta-analysis demonstrated that percentile scores as used to establish this clinical cutoff vary between clinical groups based on pain diagnoses. This brings into question the concept of a clinically relevant score: should the clinical cutoff for pain catastrophizing be based on percentiles across pain diagnoses, or is it more pertinent to establish a cutoff using comparisons with others who have a similar pain condition? Either of these options is likely to be preferable to using the current clinical cutoff based on 1 study of a sample of injured workers. Further research is necessary to establish percentile PCS scores either across or within pain conditions using raw scores from multiple studies.
Changes from baseline PCS scores following treatment such as surgery or psychological therapy were not considered in the scope of this review. However, the more precise estimates provided here for internal consistency and test–retest reliability allow calculation of reliable change beyond that attributable to random variation across time. Meta-regression following systematic review helped to refine participant variables that did (language and pain diagnosis) and did not (age and sex) relate to PCS scores over a wide population. This helps to define variables of interest for future studies of the PCS and pain catastrophizing.
Finally, the methods used in this meta-analysis could be applied to any self-report scale (PROM) used in clinical psychology or other fields. The use of meta-analysis to establish a stronger evidence base for the psychometric properties of questionnaires is encouraged to better understand sources of variance. This would strengthen the use of those self-report scales, as well as encouraging discarding those that fail standards of reliability and validity. Such research could help to introduce greater precision and theoretical justification to the ever-increasing aggregation of concepts and scales.
This is the first psychometric meta-analysis of the PCS and the first investigation of the PCS on such a large scale. Meta-analytic methods in this review confirmed the reliability of the overall scale and refined psychometric and normative properties. The PCS as a full scale is concluded to be a reliable measure. Caution is urged in the clinical interpretation of scores because of differences in scores between people with different pain diagnoses, and potential linguistic or cultural influences on PCS scores should be considered when using different language versions of the scale. Limitations are acknowledged regarding data missed by our inclusion criteria, and search terms and criteria are provided so that the review can be replicated or updated. It is hoped that results from this review will encourage the use of meta-analytic methods to establish more accurate psychometric properties of other psychological and self-report scales.
Conflict of interest statement
The authors have no conflicts of interest to declare.
The authors acknowledge Prof Robert West, Dr Gary Latchford, and Dr Andrew Prestwich for their technical help with statistical analysis and interpretation.
Supplemental digital content
Supplemental digital content associated with this article can be found online at http://links.lww.com/PAIN/A735.
. Beck AT, Rush AJ, Shaw BF, Emery G. Cognitive Therapy of Depression. New York, NY: Guilford Press, 1979.
. Bedard GBV, Reid GJ, McGrath PJ, Chambers CT. Coping and self-medication in a community sample of junior high school students. Pain Res Manag 1997;2:151–6.
. Cronbach LJ. Coefficient alpha and the internal structure of tests. Psychometrika 1951;16:297–334.
. Field A. Discovering statistics using IBM SPSS statistics. London, United Kingdom: Sage, 2013.
. Forsythe LP, Thorn B, Day M, Shelby G. Race and sex differences in primary appraisals, catastrophizing, and experimental pain outcomes. J Pain 2011;12:563–72.
. Hedges LV, Olkin I. Statistical methods for meta-analysis. Orlando: Academic Press, 1985.
. Higgins J, Thompson SG. Quantifying heterogeneity in a meta‐analysis. Stat Med 2002;21:1539–58.
. Higgins JP, Green S. Cochrane handbook for systematic reviews of interventions. Version 5.1.0 [updated March 2011]. The Cochrane Collaboration, 2011. Available at: http://handbook.cochrane.org
. Higgins JP, Thompson SG, Deeks JJ, Altman DG. Measuring inconsistency in meta-analyses. BMJ 2003;327:557–60.
. Hsieh AY, Tripp DA, Ji LJ, Sullivan MJ. Comparisons of catastrophizing, pain attitudes, and cold-pressor pain experience between Chinese and European Canadian young adults. J Pain 2010;11:1187–94.
. IBM Corp. SPSS statistics for windows. Armonk: IBM Corp, 2015.
. Jacobsen PB, Butler RW. Relation of cognitive coping and catastrophizing to acute pain and analgesic use following breast cancer surgery. J Behav Med 1996;19:17–29.
. Keefe FJ, Lefebvre JC, Egert JR, Affleck G, Sullivan MJ, Caldwell DS. The relationship of gender to pain, pain behavior, and disability in osteoarthritis patients: the role of catastrophizing. Pain 2000;87:325–34.
. Leung L. Pain catastrophizing: an updated review. Indian J Psychol Med 2012;34:204.
. Moher D, Liberati A, Tetzlaff J, Altman DG, Group P. Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement. PLoS Med 2009;6:e1000097.
. Morley S. Single-case methods in clinical psychology: A practical guide. London: Routledge, 2017.
. Osman A, Barrios FX, Gutierrez PM, Kopper BA, Merrifield T, Grittmann L. The Pain Catastrophizing Scale: further psychometric evaluation with adult samples. J Behav Med 2000;23:351–65.
. Osman A, Barrios FX, Kopper BA, Hauptmann W, Jones J, O'Neill E. Factor structure, reliability
, and validity of the pain catastrophizing scale. J Behav Med 1997;20:589–605.
. Pedler A. The Pain Catastrophising Scale. J Physiother 2010;56:137.
. Quartana PJ, Campbell CM, Edwards RR. Pain catastrophizing: a critical review. Expert Rev Neurother 2009;9:745–58.
. Riddle DL, Keefe FJ, Nay WT, McKee D, Attarian DE, Jensen MP. Pain coping skills training for patients with elevated pain catastrophizing who are scheduled for knee arthroplasty: a quasi-experimental study. Arch Phys Med Rehabil 2011;92:859–65.
. Rivest K, Côté JN, Dumas JP, Sterling M, De Serres SJ. Relationships between pain thresholds, catastrophizing and gender in acute whiplash injury. Man Ther 2010;15:154–9.
. Sullivan MJ. The pain catastrophizing scale: user manual. Montreal: McGill University, 2009. pp. 1–36.
. Sullivan MJ, Bishop SR, Pivik J. The pain catastrophizing scale: development and validation. Psychol Assess 1995;7:524.
. Sullivan MJ, Neish NR. Catastrophizing, anxiety and pain during dental hygiene treatment. Community Dent Oral Epidemiol 1998;26:344–9.
. Sullivan MJ, Thorn B, Haythornthwaite JA, Keefe F, Martin M, Bradley LA, Lefebvre JC. Theoretical perspectives on the relation between catastrophizing and pain. Clin J Pain 2001;17:52–64.
. Sullivan MJ, Tripp DA, Rodgers WM, Stanish W. Catastrophizing and pain perception in sport participants. J Appl Sport Psychol 2000;12:151–67.
. Sullivan MJ, Tripp DA, Santor D. Gender differences in pain and pain behavior: the role of catastrophizing. Cognit Ther Res 2000;24:121–34.
. Van Damme S, Crombez G, Bijttebier P, Goubert L, Van Houdenhove B. A confirmatory factor analysis of the Pain Catastrophizing Scale: invariant factor structure across clinical and non-clinical populations. PAIN 2002;96:319–24.
. Wilson DB, Lipsey MW. Practical meta-analysis. California: Sage Publications Inc, 2001.
. Wolfe F. Criteria for fibromyalgia? What is fibromyalgia? Limitations to current concepts of fibromyalgia and fibromyalgia criteria. Clin Exp Rheumatol 2017;35(suppl 105):S10–12.
Outcome measure; PROMS; Systematic review; Reliability
Supplemental Digital Content
© 2019 International Association for the Study of Pain