- Email: [email protected]
One, Two, or Three? Constructs of the Brief Pain Inventory Among Patients With Non-Cancer Pain in the Outpatient Setting
Vol. 47 No. 2 February 2014
Journal of Pain and Symptom Management
325
Brief Methodological Report
One, Two, or Three? Constructs of the Brief Pain Inventory Among Patients With Non-Cancer Pain in the Outpatient Setting Kate L. Lapane, PhD, Brian J. Quilliam, PhD, Carmela Benson, MS, Wing Chow, PharmD, MPH, and Myoung Kim, PhD Department of Quantitative Health Sciences (K.L.L.), University of Massachusetts Medical School, Worcester, Massachusetts; College of Pharmacy (B.J.Q.), University of Rhode Island, Kingston, Rhode Island; and Janssen Scientific Affairs LLC (C.B.,W.C., M.K.), Raritan, New Jersey, USA
Abstract Context. Either a two-factor representation (pain intensity and interference) or a three-factor representation (pain intensity, activity interference, and affective interference) of the modified Brief Pain Inventory (BPI) is appropriate among cancer patients. Objectives. To evaluate the extent to which a three-factor representation (pain intensity, activity interference, and affective interference) is appropriate for BPI among patients with noncancer pain seen in an outpatient setting. Methods. We conducted a prospective, multicenter, observational, nonrandomized study using patient pain registry data from outpatient settings. Seven hundred forty-one patients with acute episodes of noncancer pain requiring treatment with a prescription medication containing oxycodone immediaterelease on an as-needed basis for at least five days participated. Baseline measurements included the modified BPI pain intensity (right now, average, and worst in 24 hours) and pain interference with general activities, walking, work, mood, relations with others, sleep, and life enjoyment. Confirmatory factor analysis was conducted for the overall sample and among postoperative patients (n ¼ 133), patients with back and neck pain (n ¼ 202), patients with arthritis (n ¼ 148), and patients with injury or trauma (n ¼ 204). Results. Both the two-factor and three-factor models were statistically better than the one-factor model (P < 0.05), with the two-factor model performing better than the three-factor model. Configural invariance, but not metric invariance by patient cohort group was demonstrated. Conclusion. Consistent with analyses among cancer patients, a two-factor representation of BPI is appropriate for noncancer patients seen in an ambulatory setting. This work lends additional support for the psychometric properties of BPI. J Pain Symptom Manage 2014;47:325e333. Ó 2014 U.S. Cancer Pain Relief Committee. Published by Elsevier Inc. All rights reserved.
Address correspondence to: Kate L. Lapane, PhD, Department of Quantitative Health Science, University of Massachusetts Medical School, 55 Lake Road Ó 2014 U.S. Cancer Pain Relief Committee. Published by Elsevier Inc. All rights reserved.
North, Worcester, MA 01655, USA. [email protected] Accepted for publication: March 29, 2013.
E-mail:
0885-3924/$ - see front matter http://dx.doi.org/10.1016/j.jpainsymman.2013.03.023
326
Lapane et al.
Vol. 47 No. 2 February 2014
Key Words Pain, psychometric properties, ambulatory care, opioids, registry
Introduction The U.S. Food and Drug Administration and the European Medicines Agency have recognized the importance of patient-reported outcomes in clinical trials.1 Both regulatory agencies have released guidelines stressing the importance of content validity and construct validity of patient-reported outcome instruments.2,3 Content validity provides evidence that the instrument measures the concept of interest.4 Such evidence can come from qualitative studies that demonstrate that the items and domains of an instrument are significant and relevant, given the patient condition, patient concerns, and the instrument’s intended use.2 Construct validity provides evidence that relationships among items, domains, and concepts are consistent with a priori hypotheses about logical relationships among the concepts.4 Given the increasing importance of patient-reported outcomes and the regulatory agency guidance related to validity of instruments assessing patient-reported outcomes, studies documenting both content and construct validity of commonly used patientreported outcome instruments are needed. One important patient-reported outcome is pain. In research studies, a commonly used pain instrument is the Brief Pain Inventory (BPI).5 Although a single item is often used (pain at its worst in the last 24 hours),5 there is growing support of three relevant constructs derived from BPI: 1) pain intensity, 2) activity interference (walking, work, general activities), and 3) affective interference (relations with other people, mood, sleep, enjoyment of life).6e8 Research supporting the three constructs of BPI has been conducted in patients with cancer6e8 or HIV/AIDS.7 Yet, the extent to which these findings extend to patients with other conditions remains unknown. Using data from a comprehensive prospective, multicenter, observational, nonrandomized patient pain registry, we examined the extent to which psychometric properties regarding three distinct domains of BPI (intensity, activity interference, and affective interference) observed in a sample of patients
with cancer or HIV/AIDS extend to patients with noncancer pain being seen in an outpatient setting. By examining the construct validity of the instrument using confirmatory factor analysis, we hypothesized that two acceptable models would emerge: 1) a two-factor representation (pain intensity and interference) and 2) a three-factor representation (pain intensity, activity interference, and affective interference). Confirmatory factor analysis is a commonly used method9 to investigate construct validity.
Methods Study Sample The Oxycodone Users Registry (OUR) study,10 a prospective registry designed to provide detailed assessments of patient-reported outcomes beyond those typically available from a retrospective chart review, was used for the current analysis. Adult patients were enrolled from 48 clinical sites (providing geographic variation) during six months of 2009. Patients were eligible if their pain level required the start of a Schedule II medication containing immediate-release oxycodone (either alone or in combination with aspirin, ibuprofen, or acetaminophen) within three days after the baseline visit (or day of surgery) and to continue as needed for at least five days. Patients who reported use of a Schedule II opioid within 30 days before providing informed consent or had planned to use a Schedule II-V opioid during the study period were excluded. However, patients who switched from a Schedule III-V opioid or a non-opioid analgesic were included. Participants had pain as a result of: 1) injury or trauma of the head, neck, back, chest, or extremities; 2) fibromyalgia; 3) arthritis; 4) neuropathic pain; 5) other back or neck pain; or 6) postoperative pain after outpatient orthopedic surgery. These were typical conditions for which short-acting opioids were prescribed. Prescribers treated patients according to their usual practice with no restrictions on the dose and schedule of oxycodone. Patients used the study Web site on Days 1 (record
Vol. 47 No. 2 February 2014
Brief Pain Inventory in Non-Cancer Pain Patients
date/time of first dose of oxycodone and pain intensity at the time of dose only), 3, 7, 14, 21, and 28 (end of study) to complete assessments about patient-reported outcomes, oxycodone experience and utilization, and other medical treatments. On Day 28, the prescriber recorded final information about medical treatment and assessments of patient outcomes, with the patient chart as a reference. A total of 814 patients participated, and all provided written informed consent. We excluded patients with incomplete baseline data (n ¼ 10), as well as patients with missing data on any of the 10 pain variables used in the confirmatory factor analysis (n ¼ 27). The remaining 777 patients comprised the initial study sample. The study was approved by the New England Institutional Review Board.
Measures The modified BPI-Short Form was used to capture pain intensity and the functional aspects of the patients’ pain experience.11 Pain intensity was measured using an 11-point numeric rating scale from 0 (no pain) to 10 (worst possible pain). Patients used the 11point numeric rating scale to record pain ‘‘right now,’’ average pain in the past 24 hours, and pain at its worst in the past 24 hours. Although pain ‘‘right now’’ was thought to be free of recall error, capturing average and worst pain intensity in the past 24 hours allowed for expected diurnal variations in pain intensity. This measure has been validated in patients with chronic pain,12 as well as patients with acute pain.13 Clinically meaningful pain relief is indicated by a relative change of at least 30%e33% or an absolute change of only two points on this scale.13,14 To capture the functional aspects of the patients’ pain experience, seven additional questions used an 11-point scale (0 ¼ does not interfere, 10 ¼ completely interferes) to indicate how much pain had interfered ‘‘in the past 24 hours’’ with general activity, mood, walking ability, normal work, relations with other people, sleep, and enjoyment of life.
Statistical Analysis We used confirmatory factor analysis with maximum likelihood estimation to evaluate the extent to which previous findings7 regarding three constructs (pain intensity, activity
327
interference, and affective interference) held in noncancer ambulatory care patients. While conducting the confirmatory factor analysis, we evaluated all requirements related to normality, multicollinearity, residual values, and multivariate outliers.15 The Shapiro-Wilk test for normality revealed deviations from normality despite high values for the W statistic (>0.90). We attempted to improve normality by applying transformations (log, square, square root, 1/x, and 1/x2); however, none of these improved the performance of the variables on the Shapiro-Wilk test. Because the visual inspections of normality revealed distributions approximately normal, we included the nontransformed variables in the analysis.16 We evaluated the data for multivariate outliers that included cases with extreme scores on multiple variables or highly unusual combinations of values but no individual extreme value. We estimated Mahalanobis D to detect multivariate outliers, using P < 0.001 as a conservative significance level.17 Using this method, we identified 36 cases that were excluded from further analysis. In an iterative fashion, we examined the modification indicators (covariance), and where appropriate, treated the covariance between variables as a free parameter. When no further improvements to the model were indicated by the examination of the modification indicators, we examined the standardized residual covariance table to make sure that none exceeded an absolute value of 2. To allow for comparisons across studies, we intentionally attempted to replicate methods applied in previous work conducted in patients with HIV/AIDS or cancer.7 In line with this approach, we also used root mean squared error of approximation (RMSEA),18 comparative fit index (CFI),19 Chi-squared, and change in Chi-squared, given the change in degrees of freedom between models to test which confirmatory factor analysis model best represents the present data set. We a priori defined good models as those with an RMSEA # 0.08,20 a CFI of $0.90,20 or lower Chi-squared value, given the same number of degrees of freedom. Although developed to assess cancer pain, BPI has a theoretical foundation that indicates three domains: pain intensity, how pain interferes with activities, and how pain interferes with affect.21 Multidimensional scaling using
328
Lapane et al.
Vol. 47 No. 2 February 2014
analysis by patient cohort group: 1) postoperative patients, 2) patients with arthritis, 3) patients with back or neck pain, and 4) patients who experienced injury or trauma. We were unable to conduct the analysis in fibromyalgia or neuropathic pain owing to small sample sizes. Some analyses were conducted in SAS 9.1 (SAS Institute Inc., Cary, NC); the confirmatory factor analysis was conducted using AMOS (AMOS Development Corporation, Spring House, PA).22
samples from four countries (U.S., China, Philippines, and France)6 supports separate dimensions representing affect and activity, which were consistently interpretable across three levels of pain severity (‘‘mild,’’ ‘‘moderate,’’ and ‘‘severe’’). Based on this work as well as work by Atkinson et al.,7 we created three models: 1) a one-factor model; 2) a two-factor model with pain severity and pain interference as latent factors; and 3) a three-factor model with pain severity, pain activity interference (i.e., general activity, work, and walking ability), and affective interference (i.e., mood, relations with other people, sleep, enjoyment of life) treated as latent factors. We then conducted a multigroup analysis. First, we evaluated configural invariance to test whether the two-factor structure represented in the confirmatory factor analysis achieved adequate fit when the patient cohort groups were tested together without any cross-group path constraints. On confirmation of configural invariance, we tested for metric invariance. To test for metric invariance, we performed a Chi-squared difference test on the groups. A P-value of <0.05 was evidence of difference between the patient cohorts. Based on these results, we then stratified the
Results Table 1 shows the sociodemographic characteristics of the sample. Patients in the study were aged 18 to 84 years (mean SD: 49.8 13.1 years). Persons in the injury/ trauma cohort tended to be younger (mean: 43.9 years) than the other cohorts. Overall, 60% were women with higher representation in the neuropathic pain (69.2%) and fibromyalgia (100.0%) cohorts. Eighteen percent of participants were postoperative patients. Sixty-four percent self-reported as nonHispanic white; 11% had less than a high school education; and 54% were married.
Table 1 Sociodemographic Characteristics of Registry Participants Characteristics Age range (years) Mean Median Percentagea 18e34 35e44 45e54 55e64 65þ Women Education levelb Less than high school education High School/GED College/graduate school Other Marriedb Non-Hispanic white
Overall Postoperative Arthritis Back/Neck Injury/Trauma Neuropathic Fibromyalgia (n ¼ 741) (n ¼ 133) (n ¼ 148) (n ¼ 202) (n ¼ 204) (n ¼ 39) (n ¼ 15) 49.8 49.7
52.4 53.0
52.9 52.9
50.8 50.7
43.9 46.3
50.5 50.4
50.5 48.3
14.2 20.3 31.4 20.7 13.4 59.6
9.0 17.3 33.1 21.8 18.8 51.9
8.8 15.5 29.1 32.4 14.2 63.5
10.4 25.3 29.2 18.8 16.3 60.4
26.0 21.1 36.8 13.2 2.9 55.4
15.4 20.5 20.5 25.6 18.0 69.2
0 26.7 46.7 20.0 6.7 100.0
11.1
3.9
17.5
13.9
6.9
13.9
7.1
52.9 33.3 2.7 54.2 64.0
46.5 47.3 2.3 73.5 85.7
60.6 21.2 0.7 47.9 49.3
50.3 32.8 3.1 57.5 76.2
56.1 32.8 4.2 45.5 53.4
55.6 30.6 0.0 42.1 56.4
71.4 21.4 0.0 42.9 60.0
8.55 (1.42) 7.12 (1.71) 7.03 (1.89)
8.74 (1.55) 7.77 (1.78) 7.23 (2.33)
8.54 (2.0) 8.13 (1.85) 8.27 (1.44)
Pain scores Worst in 24 hours Average Right now
Mean (SD) 8.00 (2.11) 6.65 (2.16) 6.36 (2.38)
6.26 (2.93) 4.58 (2.65) 4.16 (2.55)
GED ¼ general equivalency diploma. a May not total 100% owing to rounding. b Missing data: education (n ¼ 41) and marital status (n ¼ 11).
7.99 (2.09) 8.42 (1.51) 6.89 (1.84) 7.04 (1.65) 6.56 (2.15) 6.67 (2.05)
Vol. 47 No. 2 February 2014
Brief Pain Inventory in Non-Cancer Pain Patients
329
the overall sample, as well as patient cohorts. Table 4 shows some variability in the factor loading by patient cohort. For example, for the back/neck cohort and the injury/trauma cohorts, the factor loadings for interference with sleep, relations with other people, and walking ability hovered just below 0.70.
The mean pain scores are shown by group. The pain scores in the postoperative group are lower than those in the other cohorts. Table 2 provides the means and SDs for individual variables, as well as the Pearson correlation coefficients among BPI variables included in this study. All correlations shown in Table 2 were statistically significant (P < 0.0001). We also conducted the analysis on Table 2 data stratified by patient cohort. In all patient cohorts, the correlations remained statistically significant (P < 0.0001). For the most part, however, the correlations were higher among the postoperative cohort relative to the patients with other indications for pain. Table 3 shows the fit indices for three models: unitary, two factor, and three factor. Relative to the one-factor model in the overall sample, Model 2 (representing two domains: pain and interference) provided a lower RMSEA (0.05) and a higher CFI (0.99) and a significant change in Chi-squared value, given the change in degrees of freedom when compared with Model 1 (P < 0.05). Model 3 did not appear to offer statistical superiority. We show Model 2 as the best fit for the data (Fig. 1). We then conducted a multigroup analysis by cohort type (postoperative, back/neck, arthritis, and injury/trauma). Although this analysis revealed configural invariance (PCLOSE ¼ 0.112; RMSEA ¼ 0.05; CFI ¼ 0.93), it did not reveal metric invariance. An analysis of the paths revealed that the patient cohorts differed on the three pain measures, general activities, work, and sleep. Table 4 provides the standardized factor loadings derived from the two-factor models for
Discussion Our analysis of data from the OUR Study provided the ability to evaluate the extent to which a two-factor or a three-factor construct is appropriate for BPI among patients in an ambulatory setting with arthritis, pain from injury/trauma, back or neck pain, and postoperative pain. Research supporting the use of BPI in noncancer patients has included patients with HIV/ AIDS,23 postsurgical patients,24,25 and persons with either low back pain or arthritis.11 Whereas recent research in patients with HIV/AIDS or cancer provides support for a two-factor (pain intensity and activity interference) or a threefactor model (pain intensity, activity interference, and affective interference), our study provides support for a two-factor representation (pain intensity and activity interference) of BPI in noncancer patients. The population included in our study contains similar groups assessed using BPI (i.e., osteoarthritis,26 neuropathy27). However, the OUR Study also included patients experiencing sources of pain commonly encountered in everyday clinical practice including arthritis, back/neck pain, injury-/trauma-related pain, neuropathic pain, and fibromyalgia-related
Table 2 Correlation Coefficients, Means, and SDs for Outcome Measures (n ¼ 741) Variables 1. Pain at its worst in the last 24 hours 2. Pain on average 3. Pain right now 4. Interference with general activity 5. Interference with walking ability 6. Interference with work 7. Interference with mood 8. Interference with relations/other people 9. Interference with sleep 10. Interference with enjoyment of life Mean SD
1
2
3
4
5
6
7
8
9
10
1.0 0.77 0.69 0.70 0.50 0.58 0.63 0.51 0.62 0.53 8.00 2.11
1.0 0.79 0.72 0.54 0.61 0.61 0.50 0.64 0.56 6.65 2.16
1.0 0.68 0.47 0.53 0.55 0.46 0.57 0.51 6.36 2.38
1.0 0.57 0.72 0.71 0.58 0.64 0.66 6.65 2.50
1.0 0.68 0.59 0.63 0.54 0.60 5.99 3.12
1.0 0.70 0.62 0.64 0.73 6.75 2.68
1.0 0.76 0.66 0.68 5.89 2.81
1.0 0.59 0.68 4.59 3.15
1.0 0.68 6.58 2.85
1.0 6.52 2.77
All correlations were statistically significant, P < 0.0001.
330
Lapane et al.
Vol. 47 No. 2 February 2014
Table 3 Fit Indices for Confirmatory Factor Models in Overall Sample Model
RMSEA
Overall (n ¼ 741)a Model 1: unitary Model 2: pain intensity, Model 3: pain intensity, Postoperative (n ¼ 133) Model 1: unitary Model 2: pain intensity, Model 3: pain intensity, Arthritis (n ¼ 148) Model 1: unitary Model 2: pain intensity, Model 3: pain intensity, Back/neck (n ¼ 202) Model 1: unitary Model 2: pain intensity, Model 3: pain intensity, Injury/trauma (n ¼ 204) Model 1: unitary Model 2: pain intensity, Model 3: pain intensity,
90% CI
CFI
df
X2
X2/df
P
interference activity interference, affective interference
0.17 0.05 0.12
0.16e0.18 0.87 35 803.33 0.04e0.07 0.99 20 60.08 0.09e0.14 0.95 30 88.04
3.00 2.94
<0.05 <0.05
interference activity interference, affective interference
0.18 0.07 0.11
0.16e0.21 0.87 35 189.13 0.01e0.10 0.99 22 34.20 0.08e0.14 0.96 29 73.41
1.55 2.53
0.05 <0.05
interference activity interference, affective interference
0.18 0.11 0.12
0.15e0.20 0.86 35 198.34 0.07e0.13 0.96 29 76.03 0.09e0.14 0.95 30 88.04
2.62 2.94
<0.05 <0.05
interference activity interference, affective interference
0.15 0.08 0.07
0.13e0.17 0.86 35 189.75 0.06e0.10 0.96 32 72.93 0.05e0.10 0.97 30 61.10
2.26 2.04
<0.05 <0.05
interference activity interference, affective interference
0.17 0.08 0.09
0.15e0.19 0.82 35 235.78 0.05e0.10 0.97 30 2.17 0.06e0.11 0.96 30 2.47
65.15 74.08
<0.05 <0.05
RMSEA ¼ root mean square error of approximation; CFI ¼ comparative fit index; df ¼ degrees of freedom. a We excluded 36 observations that were multivariate outliers.
pain. For these groups, the construct validity of the modified BPI has not (to our knowledge) been well studied. In addition, our population included persons managed for postoperative pain in the outpatient setting. In a 2006 randomized trial of patients with osteoarthritis,28 the modified short form of BPI exhibited a three-factor structure (pain intensity, activity interference, and affective interference). Yet, recommendations for using the standard short-form BPI suggested a two-factor structure
including pain intensity and activity interference.29 Our findings indicate that a two-factor structure (pain intensity and interference) performed better than did the three-factor model, suggesting the two-factor model may be appropriate for noncancer pain associated with medical conditions physicians see routinely in daily clinical practice. The diversity of the patient cohorts with respect to the underlying cause of pain must be considered. Fear related to pain intensity
Fig. 1. Confirmatory factor model for the two-factor solutiondoverall.
Vol. 47 No. 2 February 2014
Brief Pain Inventory in Non-Cancer Pain Patients
331
Table 4 Standardized Factor Loadings and Factor Correlations by Patient Cohort and Latent Construct for Two-Factor Model Variables Standardized factor coefficient Factor 1: pain intensity Pain at its worst in the last 24 hours Pain on average Pain right now Factor 2: interference Interference with general activity Interference with walking ability Interference with work Interference with mood Interference with relations/other people Interference with sleep Interference with enjoyment of life Factor correlations Factor correlation
Overalla (n ¼ 741)
Postoperative (n ¼ 133)
Arthritis (n ¼ 148)
Back/Neck (n ¼ 202)
Injury/Trauma (n ¼ 204)
0.87 0.88 0.81
0.93 0.95 0.88
0.73 0.89 0.79
0.74 0.86 0.76
0.79 0.84 0.81
0.95 0.66 0.77 0.88 0.71 0.85 0.80
0.91 d 0.84 0.85 0.71 0.85 0.82
0.86 0.81 0.86 0.81 0.72 0.79 0.86
0.80 0.68 0.77 0.77 0.67 0.67 0.79
0.78 0.69 0.80 0.79 0.68 0.66 0.80
0.84
0.88
0.83
0.79
0.68
a
Includes patients with fibromyalgia and neuropathic pain.
may lead to behavior modification.30 This suggests that pain may cause patients to change their activities both directly (because of pain experienced during the activity) and indirectly (because of fear of pain that might be experienced during the activity). Although the motivation for interference with activity was beyond the scope of our study, our findings suggest that both pain intensity and activity interference are important factors to consider when evaluating pain if including both acute (i.e., postoperative pain) and acute exacerbation of chronic pain (i.e., arthritis) populations. Our findings suggest that different aspects of each of the factors were more or less important within each of the subgroup populations we studied, most notable between the postoperative and other chronic pain conditions. Our results concur with previous research11 supporting the notion that BPI activity interference also may be useful to assess medical outcomes. Further assessment of the differential contribution of items within each of the factors may be an important avenue for future research, specifically focusing on elucidating differences between acute and chronic pain. The multigroup analysis we conducted supported the same two-factor configuration, regardless of patient cohort type; however, it did not support metric invariance across the patient cohort type. A path analysis revealed that differences across the patient cohorts were apparent for the three pain measures, general activities, work, and sleep. This is not
completely unexpected, as our patient cohort types included both individuals experiencing acute exacerbation of chronic pain (e.g., neuropathic pain, arthritis, and fibromyalgia) and patient cohort types experiencing acute experiences of pain that is expected to be of relatively short duration (e.g., postoperative pain, injury/trauma). Indeed, an analysis of the means and SDs of pain measures show some variation across groups. With this in mind, perception of pain is likely to vary in situations. For example, a 2004 study by Grotle et al.31 assessed differences in fear avoidance between persons with acute low back pain and a second group with chronic back pain. Lower levels of fear avoidance were observed in the population with acute back pain compared with those with chronic back pain, even after controlling for depression.32 Catastrophizing, attaching negative emotions to pain,32 has been theorized to modify the pain experience. A systematic literature review of catastrophizing behavior in arthritis, fibromyalgia, and other related diseases concluded that catastrophizing is related to pain and pain management-related outcomes.33 These and other studies suggest that relationships between pain assessment and the patient’s pain experience can vary across different acute and chronic conditions that may be associated with pain. The changing regulatory environment has placed increasing importance on the use of patient-reported outcomes in clinical trials,1
332
Lapane et al.
with clear guidance regarding the need for better understanding of the psychometric properties of instruments as used in different patient cohorts.2,3 Additional information regarding the construct validity of the modified BPI in specific patient populations is warranted. BPI is a relevant instrument for use in research for several reasons. First, completing BPI requires little time, and as such, it has been recommended in studies requiring repeated assessment over the course of the study. BPI captures the relevant information needed, with relatively few questions. Furthermore, BPI has been shown to be responsive to change over time in behavioral34 and pharmacological35 interventions. Our findings must be considered with several caveats in mind. The modified BPI includes three measures of pain intensity. As such, the findings of this study may not be generalizable to studies in which BPI with four measures of pain intensity were captured. Although the registry set out to capture at least 200 patients in each of six patient cohorts, recruitment goals were not met in several of the cohorts. Therefore, we did not have a sufficient sample size to evaluate BPI in patients with fibromyalgia and patients with neuropathic pain.
Conclusion Our study was conducted in an outpatient setting with patients experiencing noncancer pain and provides further evidence of the psychometric properties of BPI. We have confirmed the construct validity of BPI in ambulatory patients. Despite the differences in the patient population, our findings reflected the work of others. That is, the confirmatory factor analysis supports the two-factor representation of BPI. As selecting the appropriate instruments is essential for the robustness of any study, this study provides additional support in favor of BPI.
Disclosures and Acknowledgments This work was funded through a consulting agreement with Janssen Scientific Affairs, LLC to Dr. Lapane and Dr. Quilliam. The authors declare no additional potential conflicts of interest.
Vol. 47 No. 2 February 2014
References 1. Bottomley A, Jones D, Claassens L. Patient-reported outcomes: assessment and current perspectives of the guidelines of the Food and Drug Administration and the reflection paper of the European Medicines Agency. Eur J Cancer 2009;45: 347e353. 2. U.S. Food and Drug Administration. Guidance for industry. Patient-reported outcome measures: use in medical development to support labeling claims. 2009. Available from http://www.fda.gov/ downloads/Drugs/GuidanceComplianceRegulatory Information/Guidances/UCM193282.pdf. Accessed March 4, 2012. 3. European Medicines Agency. Committee for Medicinal Products for Human Use (CHMP). Reflection paper on the regulatory guidance for the use of health-related quality of life (HRQL) measures in the evaluation of medicinal products. 2005. Available from http://www.emea.europa.eu/pdfs/human/ ewp/13939104en.pdf. Accessed March 4, 2012. 4. Streiner DL, Norman GR. Health measurement scales. A practical guide to their development and use, 2nd ed. New York: Oxford Medical Publications, Inc., 1995. 5. Atkinson TM, Mendoza TR, Sit L, et al. The Brief Pain Inventory and its ‘‘pain at its worst in the last 24 hours’’ item: clinical trial endpoint considerations. Pain Med 2010;11:337e346. 6. Cleeland CS, Nakamura Y, Mendoza TR, et al. Dimensions of the impact of cancer pain in a four country sample: new information from multidimensional scaling. Pain 1996;67:267e273. 7. Atkinson TM, Rosenfeld BD, Sit L, et al. Using confirmatory factor analysis to evaluate construct validity of the Brief Pain Inventory (BPI). J Pain Symptom Manage 2011;41:558e565. 8. Atkinson TM, Halabi S, Bennett AV, et al. Cancer and Leukemia Group B. Measurement of affective and activity pain interference using the Brief Pain Inventory (BPI): Cancer and Leukemia Group B 70903. Pain Med 2012;13:1417e1424. 9. Brown TA. Confirmatory factor analysis for applied research. New York: Guilford, 2006. 10. Kim M, Chow W, Benson C. Rationale and design of the Oxycodone Users Registry: A prospective, multicenter registry of patients with nonmalignant pain. J Opioid Manag 2013;9(3): 189e204. 11. Keller S, Bann CM, Dodd SL, et al. Validity of the Brief Pain Inventory for use in documenting the outcomes of patients with noncancer pain. Clin J Pain 2004;20:309e318. 12. Farrar JT, Young JP Jr, LaMoreaux L, Werth JL, Poole RM. Clinical importance of changes in chronic pain intensity measured on an 11-point numerical pain rating scale. Pain 2001;94:149e158.
Vol. 47 No. 2 February 2014
Brief Pain Inventory in Non-Cancer Pain Patients
13. Farrar JT, Berlin JA, Strom BL. Clinically important changes in acute pain outcome measures: a validation study. J Pain Symptom Manage 2003;25: 406e411. 14. Farrar JT, Portenoy RK, Berlin JA, Kinman JL, Strom BL. Defining the clinically important difference in pain outcome measures. Pain 2000;88: 287e294. 15. Schreiber JB, Stage FK, King J, Nora A, Barlow EA. Reporting structural equation modeling and confirmatory factor analysis results: a review. J Educ Res 2006;99:323e337. 16. Bollen KA. Structural equations with latent variables, 2nd ed. New York: Wiley, 1995. 17. Kline RB. Principles and practice of structural equation modeling, 2nd ed. New York: Guilford, 2005. 18. Browne MW, Cudeck R. Alternative ways of assessing model fit. Newbury Park, CA: Sage, 1993. 19. Bentler PM. Comparative fit indexes in structural models. Psychol Bull 1990;107:238e246. 20. Hu L, Bentler PM. Cutoff criteria for fit indexes in covariance structure analysis: conventional criteria versus new alternatives. Struct Equation Model 1999;6:1e55. 21. Daut RL, Cleeland CS. The prevalence and severity of pain in cancer. Cancer 1982;50:1913e1918. 22. Amos [computer program], version 20. Meadville, PA: AMOS Develpoment Corporation, 2012. 23. Breitbart W, McDonald MV, Rosenfeld B, et al. Pain in ambulatory AIDS patients. I: Pain characteristics and medical correlates. Pain 1996;68:315e321. 24. Beauregard L, Pomp A, Choiniere M. Severity and impact of pain after day surgery. Can J Anaesth 1998;45:304e311. 25. Zalon ML. Comparison of pain measures in surgical patients. J Nurs Meas 1999;7:135e152. 26. Roth SH, Fleischmann RM, Burch FX, et al. Around-the-clock, controlled-release oxycodone
333
therapy for osteoarthritis-related pain: placebocontrolled trial and long-term evaluation. Arch Intern Med 2000;160:853e860. 27. Semenchuk MR, Sherman S, Davis B. Doubleblind, randomized trial of bupropion SR for the treatment of neuropathic pain. Neurology 2001; 57:1583e1588. 28. Mendoza T, Mayne T, Rublee D, Cleeland C. Reliability and validity of a modified Brief Pain Inventory short form in patients with osteoarthritis. Eur J Pain 2005;10:353e361. 29. Turk DC, Dworkin RH, Allen RR, et al. Core outcome domains for chronic pain clinical trials: IMMPACT recommendations. Pain 2003;106: 337e345. 30. Vlaeyen JWS, Kole-Snijders AMJ, Boren RGB, VanEek H. Fear of movement/(re)injury in chronic low back pain and its relation to behavioral performance. Pain 1995;62:363e372. 31. Grotle M, Vollestad NK, Veierod MB, Brox JI. Fear-avoidance beliefs and distress in relation to disability in acute and chronic low back pain. Pain 2004;112:343e352. 32. Sullivan MJ, Thorn B, Haythornthwaite JA, et al. Theoretical perspectives on the relation between catastrophizing and pain. Clin J Pain 2001;17: 52e64. 33. Edwards RR, Bingham CO, Bathon J, Haythornthwaite JA. Catastrophizing and pain in arthritis, fibromyalgia, and other rheumatic diseases. Arthritis Rheum 2006;55:325e332. 34. Twycross R, Harcourt J, Bergl S. A survey of pain in patients with advanced cancer. J Pain Symptom Manage 1996;12:273e282. 35. Hwang SS, Chang VT, Kasimis B. Dynamic cancer pain management outcomes: the relationship between pain severity, pain relief, functional interference, satisfaction and global quality of life over time. J Pain Symptom Manage 2002;23:190e200.
Journal of Pain and Symptom Management
325
Brief Methodological Report
One, Two, or Three? Constructs of the Brief Pain Inventory Among Patients With Non-Cancer Pain in the Outpatient Setting Kate L. Lapane, PhD, Brian J. Quilliam, PhD, Carmela Benson, MS, Wing Chow, PharmD, MPH, and Myoung Kim, PhD Department of Quantitative Health Sciences (K.L.L.), University of Massachusetts Medical School, Worcester, Massachusetts; College of Pharmacy (B.J.Q.), University of Rhode Island, Kingston, Rhode Island; and Janssen Scientific Affairs LLC (C.B.,W.C., M.K.), Raritan, New Jersey, USA
Abstract Context. Either a two-factor representation (pain intensity and interference) or a three-factor representation (pain intensity, activity interference, and affective interference) of the modified Brief Pain Inventory (BPI) is appropriate among cancer patients. Objectives. To evaluate the extent to which a three-factor representation (pain intensity, activity interference, and affective interference) is appropriate for BPI among patients with noncancer pain seen in an outpatient setting. Methods. We conducted a prospective, multicenter, observational, nonrandomized study using patient pain registry data from outpatient settings. Seven hundred forty-one patients with acute episodes of noncancer pain requiring treatment with a prescription medication containing oxycodone immediaterelease on an as-needed basis for at least five days participated. Baseline measurements included the modified BPI pain intensity (right now, average, and worst in 24 hours) and pain interference with general activities, walking, work, mood, relations with others, sleep, and life enjoyment. Confirmatory factor analysis was conducted for the overall sample and among postoperative patients (n ¼ 133), patients with back and neck pain (n ¼ 202), patients with arthritis (n ¼ 148), and patients with injury or trauma (n ¼ 204). Results. Both the two-factor and three-factor models were statistically better than the one-factor model (P < 0.05), with the two-factor model performing better than the three-factor model. Configural invariance, but not metric invariance by patient cohort group was demonstrated. Conclusion. Consistent with analyses among cancer patients, a two-factor representation of BPI is appropriate for noncancer patients seen in an ambulatory setting. This work lends additional support for the psychometric properties of BPI. J Pain Symptom Manage 2014;47:325e333. Ó 2014 U.S. Cancer Pain Relief Committee. Published by Elsevier Inc. All rights reserved.
Address correspondence to: Kate L. Lapane, PhD, Department of Quantitative Health Science, University of Massachusetts Medical School, 55 Lake Road Ó 2014 U.S. Cancer Pain Relief Committee. Published by Elsevier Inc. All rights reserved.
North, Worcester, MA 01655, USA. [email protected] Accepted for publication: March 29, 2013.
E-mail:
0885-3924/$ - see front matter http://dx.doi.org/10.1016/j.jpainsymman.2013.03.023
326
Lapane et al.
Vol. 47 No. 2 February 2014
Key Words Pain, psychometric properties, ambulatory care, opioids, registry
Introduction The U.S. Food and Drug Administration and the European Medicines Agency have recognized the importance of patient-reported outcomes in clinical trials.1 Both regulatory agencies have released guidelines stressing the importance of content validity and construct validity of patient-reported outcome instruments.2,3 Content validity provides evidence that the instrument measures the concept of interest.4 Such evidence can come from qualitative studies that demonstrate that the items and domains of an instrument are significant and relevant, given the patient condition, patient concerns, and the instrument’s intended use.2 Construct validity provides evidence that relationships among items, domains, and concepts are consistent with a priori hypotheses about logical relationships among the concepts.4 Given the increasing importance of patient-reported outcomes and the regulatory agency guidance related to validity of instruments assessing patient-reported outcomes, studies documenting both content and construct validity of commonly used patientreported outcome instruments are needed. One important patient-reported outcome is pain. In research studies, a commonly used pain instrument is the Brief Pain Inventory (BPI).5 Although a single item is often used (pain at its worst in the last 24 hours),5 there is growing support of three relevant constructs derived from BPI: 1) pain intensity, 2) activity interference (walking, work, general activities), and 3) affective interference (relations with other people, mood, sleep, enjoyment of life).6e8 Research supporting the three constructs of BPI has been conducted in patients with cancer6e8 or HIV/AIDS.7 Yet, the extent to which these findings extend to patients with other conditions remains unknown. Using data from a comprehensive prospective, multicenter, observational, nonrandomized patient pain registry, we examined the extent to which psychometric properties regarding three distinct domains of BPI (intensity, activity interference, and affective interference) observed in a sample of patients
with cancer or HIV/AIDS extend to patients with noncancer pain being seen in an outpatient setting. By examining the construct validity of the instrument using confirmatory factor analysis, we hypothesized that two acceptable models would emerge: 1) a two-factor representation (pain intensity and interference) and 2) a three-factor representation (pain intensity, activity interference, and affective interference). Confirmatory factor analysis is a commonly used method9 to investigate construct validity.
Methods Study Sample The Oxycodone Users Registry (OUR) study,10 a prospective registry designed to provide detailed assessments of patient-reported outcomes beyond those typically available from a retrospective chart review, was used for the current analysis. Adult patients were enrolled from 48 clinical sites (providing geographic variation) during six months of 2009. Patients were eligible if their pain level required the start of a Schedule II medication containing immediate-release oxycodone (either alone or in combination with aspirin, ibuprofen, or acetaminophen) within three days after the baseline visit (or day of surgery) and to continue as needed for at least five days. Patients who reported use of a Schedule II opioid within 30 days before providing informed consent or had planned to use a Schedule II-V opioid during the study period were excluded. However, patients who switched from a Schedule III-V opioid or a non-opioid analgesic were included. Participants had pain as a result of: 1) injury or trauma of the head, neck, back, chest, or extremities; 2) fibromyalgia; 3) arthritis; 4) neuropathic pain; 5) other back or neck pain; or 6) postoperative pain after outpatient orthopedic surgery. These were typical conditions for which short-acting opioids were prescribed. Prescribers treated patients according to their usual practice with no restrictions on the dose and schedule of oxycodone. Patients used the study Web site on Days 1 (record
Vol. 47 No. 2 February 2014
Brief Pain Inventory in Non-Cancer Pain Patients
date/time of first dose of oxycodone and pain intensity at the time of dose only), 3, 7, 14, 21, and 28 (end of study) to complete assessments about patient-reported outcomes, oxycodone experience and utilization, and other medical treatments. On Day 28, the prescriber recorded final information about medical treatment and assessments of patient outcomes, with the patient chart as a reference. A total of 814 patients participated, and all provided written informed consent. We excluded patients with incomplete baseline data (n ¼ 10), as well as patients with missing data on any of the 10 pain variables used in the confirmatory factor analysis (n ¼ 27). The remaining 777 patients comprised the initial study sample. The study was approved by the New England Institutional Review Board.
Measures The modified BPI-Short Form was used to capture pain intensity and the functional aspects of the patients’ pain experience.11 Pain intensity was measured using an 11-point numeric rating scale from 0 (no pain) to 10 (worst possible pain). Patients used the 11point numeric rating scale to record pain ‘‘right now,’’ average pain in the past 24 hours, and pain at its worst in the past 24 hours. Although pain ‘‘right now’’ was thought to be free of recall error, capturing average and worst pain intensity in the past 24 hours allowed for expected diurnal variations in pain intensity. This measure has been validated in patients with chronic pain,12 as well as patients with acute pain.13 Clinically meaningful pain relief is indicated by a relative change of at least 30%e33% or an absolute change of only two points on this scale.13,14 To capture the functional aspects of the patients’ pain experience, seven additional questions used an 11-point scale (0 ¼ does not interfere, 10 ¼ completely interferes) to indicate how much pain had interfered ‘‘in the past 24 hours’’ with general activity, mood, walking ability, normal work, relations with other people, sleep, and enjoyment of life.
Statistical Analysis We used confirmatory factor analysis with maximum likelihood estimation to evaluate the extent to which previous findings7 regarding three constructs (pain intensity, activity
327
interference, and affective interference) held in noncancer ambulatory care patients. While conducting the confirmatory factor analysis, we evaluated all requirements related to normality, multicollinearity, residual values, and multivariate outliers.15 The Shapiro-Wilk test for normality revealed deviations from normality despite high values for the W statistic (>0.90). We attempted to improve normality by applying transformations (log, square, square root, 1/x, and 1/x2); however, none of these improved the performance of the variables on the Shapiro-Wilk test. Because the visual inspections of normality revealed distributions approximately normal, we included the nontransformed variables in the analysis.16 We evaluated the data for multivariate outliers that included cases with extreme scores on multiple variables or highly unusual combinations of values but no individual extreme value. We estimated Mahalanobis D to detect multivariate outliers, using P < 0.001 as a conservative significance level.17 Using this method, we identified 36 cases that were excluded from further analysis. In an iterative fashion, we examined the modification indicators (covariance), and where appropriate, treated the covariance between variables as a free parameter. When no further improvements to the model were indicated by the examination of the modification indicators, we examined the standardized residual covariance table to make sure that none exceeded an absolute value of 2. To allow for comparisons across studies, we intentionally attempted to replicate methods applied in previous work conducted in patients with HIV/AIDS or cancer.7 In line with this approach, we also used root mean squared error of approximation (RMSEA),18 comparative fit index (CFI),19 Chi-squared, and change in Chi-squared, given the change in degrees of freedom between models to test which confirmatory factor analysis model best represents the present data set. We a priori defined good models as those with an RMSEA # 0.08,20 a CFI of $0.90,20 or lower Chi-squared value, given the same number of degrees of freedom. Although developed to assess cancer pain, BPI has a theoretical foundation that indicates three domains: pain intensity, how pain interferes with activities, and how pain interferes with affect.21 Multidimensional scaling using
328
Lapane et al.
Vol. 47 No. 2 February 2014
analysis by patient cohort group: 1) postoperative patients, 2) patients with arthritis, 3) patients with back or neck pain, and 4) patients who experienced injury or trauma. We were unable to conduct the analysis in fibromyalgia or neuropathic pain owing to small sample sizes. Some analyses were conducted in SAS 9.1 (SAS Institute Inc., Cary, NC); the confirmatory factor analysis was conducted using AMOS (AMOS Development Corporation, Spring House, PA).22
samples from four countries (U.S., China, Philippines, and France)6 supports separate dimensions representing affect and activity, which were consistently interpretable across three levels of pain severity (‘‘mild,’’ ‘‘moderate,’’ and ‘‘severe’’). Based on this work as well as work by Atkinson et al.,7 we created three models: 1) a one-factor model; 2) a two-factor model with pain severity and pain interference as latent factors; and 3) a three-factor model with pain severity, pain activity interference (i.e., general activity, work, and walking ability), and affective interference (i.e., mood, relations with other people, sleep, enjoyment of life) treated as latent factors. We then conducted a multigroup analysis. First, we evaluated configural invariance to test whether the two-factor structure represented in the confirmatory factor analysis achieved adequate fit when the patient cohort groups were tested together without any cross-group path constraints. On confirmation of configural invariance, we tested for metric invariance. To test for metric invariance, we performed a Chi-squared difference test on the groups. A P-value of <0.05 was evidence of difference between the patient cohorts. Based on these results, we then stratified the
Results Table 1 shows the sociodemographic characteristics of the sample. Patients in the study were aged 18 to 84 years (mean SD: 49.8 13.1 years). Persons in the injury/ trauma cohort tended to be younger (mean: 43.9 years) than the other cohorts. Overall, 60% were women with higher representation in the neuropathic pain (69.2%) and fibromyalgia (100.0%) cohorts. Eighteen percent of participants were postoperative patients. Sixty-four percent self-reported as nonHispanic white; 11% had less than a high school education; and 54% were married.
Table 1 Sociodemographic Characteristics of Registry Participants Characteristics Age range (years) Mean Median Percentagea 18e34 35e44 45e54 55e64 65þ Women Education levelb Less than high school education High School/GED College/graduate school Other Marriedb Non-Hispanic white
Overall Postoperative Arthritis Back/Neck Injury/Trauma Neuropathic Fibromyalgia (n ¼ 741) (n ¼ 133) (n ¼ 148) (n ¼ 202) (n ¼ 204) (n ¼ 39) (n ¼ 15) 49.8 49.7
52.4 53.0
52.9 52.9
50.8 50.7
43.9 46.3
50.5 50.4
50.5 48.3
14.2 20.3 31.4 20.7 13.4 59.6
9.0 17.3 33.1 21.8 18.8 51.9
8.8 15.5 29.1 32.4 14.2 63.5
10.4 25.3 29.2 18.8 16.3 60.4
26.0 21.1 36.8 13.2 2.9 55.4
15.4 20.5 20.5 25.6 18.0 69.2
0 26.7 46.7 20.0 6.7 100.0
11.1
3.9
17.5
13.9
6.9
13.9
7.1
52.9 33.3 2.7 54.2 64.0
46.5 47.3 2.3 73.5 85.7
60.6 21.2 0.7 47.9 49.3
50.3 32.8 3.1 57.5 76.2
56.1 32.8 4.2 45.5 53.4
55.6 30.6 0.0 42.1 56.4
71.4 21.4 0.0 42.9 60.0
8.55 (1.42) 7.12 (1.71) 7.03 (1.89)
8.74 (1.55) 7.77 (1.78) 7.23 (2.33)
8.54 (2.0) 8.13 (1.85) 8.27 (1.44)
Pain scores Worst in 24 hours Average Right now
Mean (SD) 8.00 (2.11) 6.65 (2.16) 6.36 (2.38)
6.26 (2.93) 4.58 (2.65) 4.16 (2.55)
GED ¼ general equivalency diploma. a May not total 100% owing to rounding. b Missing data: education (n ¼ 41) and marital status (n ¼ 11).
7.99 (2.09) 8.42 (1.51) 6.89 (1.84) 7.04 (1.65) 6.56 (2.15) 6.67 (2.05)
Vol. 47 No. 2 February 2014
Brief Pain Inventory in Non-Cancer Pain Patients
329
the overall sample, as well as patient cohorts. Table 4 shows some variability in the factor loading by patient cohort. For example, for the back/neck cohort and the injury/trauma cohorts, the factor loadings for interference with sleep, relations with other people, and walking ability hovered just below 0.70.
The mean pain scores are shown by group. The pain scores in the postoperative group are lower than those in the other cohorts. Table 2 provides the means and SDs for individual variables, as well as the Pearson correlation coefficients among BPI variables included in this study. All correlations shown in Table 2 were statistically significant (P < 0.0001). We also conducted the analysis on Table 2 data stratified by patient cohort. In all patient cohorts, the correlations remained statistically significant (P < 0.0001). For the most part, however, the correlations were higher among the postoperative cohort relative to the patients with other indications for pain. Table 3 shows the fit indices for three models: unitary, two factor, and three factor. Relative to the one-factor model in the overall sample, Model 2 (representing two domains: pain and interference) provided a lower RMSEA (0.05) and a higher CFI (0.99) and a significant change in Chi-squared value, given the change in degrees of freedom when compared with Model 1 (P < 0.05). Model 3 did not appear to offer statistical superiority. We show Model 2 as the best fit for the data (Fig. 1). We then conducted a multigroup analysis by cohort type (postoperative, back/neck, arthritis, and injury/trauma). Although this analysis revealed configural invariance (PCLOSE ¼ 0.112; RMSEA ¼ 0.05; CFI ¼ 0.93), it did not reveal metric invariance. An analysis of the paths revealed that the patient cohorts differed on the three pain measures, general activities, work, and sleep. Table 4 provides the standardized factor loadings derived from the two-factor models for
Discussion Our analysis of data from the OUR Study provided the ability to evaluate the extent to which a two-factor or a three-factor construct is appropriate for BPI among patients in an ambulatory setting with arthritis, pain from injury/trauma, back or neck pain, and postoperative pain. Research supporting the use of BPI in noncancer patients has included patients with HIV/ AIDS,23 postsurgical patients,24,25 and persons with either low back pain or arthritis.11 Whereas recent research in patients with HIV/AIDS or cancer provides support for a two-factor (pain intensity and activity interference) or a threefactor model (pain intensity, activity interference, and affective interference), our study provides support for a two-factor representation (pain intensity and activity interference) of BPI in noncancer patients. The population included in our study contains similar groups assessed using BPI (i.e., osteoarthritis,26 neuropathy27). However, the OUR Study also included patients experiencing sources of pain commonly encountered in everyday clinical practice including arthritis, back/neck pain, injury-/trauma-related pain, neuropathic pain, and fibromyalgia-related
Table 2 Correlation Coefficients, Means, and SDs for Outcome Measures (n ¼ 741) Variables 1. Pain at its worst in the last 24 hours 2. Pain on average 3. Pain right now 4. Interference with general activity 5. Interference with walking ability 6. Interference with work 7. Interference with mood 8. Interference with relations/other people 9. Interference with sleep 10. Interference with enjoyment of life Mean SD
1
2
3
4
5
6
7
8
9
10
1.0 0.77 0.69 0.70 0.50 0.58 0.63 0.51 0.62 0.53 8.00 2.11
1.0 0.79 0.72 0.54 0.61 0.61 0.50 0.64 0.56 6.65 2.16
1.0 0.68 0.47 0.53 0.55 0.46 0.57 0.51 6.36 2.38
1.0 0.57 0.72 0.71 0.58 0.64 0.66 6.65 2.50
1.0 0.68 0.59 0.63 0.54 0.60 5.99 3.12
1.0 0.70 0.62 0.64 0.73 6.75 2.68
1.0 0.76 0.66 0.68 5.89 2.81
1.0 0.59 0.68 4.59 3.15
1.0 0.68 6.58 2.85
1.0 6.52 2.77
All correlations were statistically significant, P < 0.0001.
330
Lapane et al.
Vol. 47 No. 2 February 2014
Table 3 Fit Indices for Confirmatory Factor Models in Overall Sample Model
RMSEA
Overall (n ¼ 741)a Model 1: unitary Model 2: pain intensity, Model 3: pain intensity, Postoperative (n ¼ 133) Model 1: unitary Model 2: pain intensity, Model 3: pain intensity, Arthritis (n ¼ 148) Model 1: unitary Model 2: pain intensity, Model 3: pain intensity, Back/neck (n ¼ 202) Model 1: unitary Model 2: pain intensity, Model 3: pain intensity, Injury/trauma (n ¼ 204) Model 1: unitary Model 2: pain intensity, Model 3: pain intensity,
90% CI
CFI
df
X2
X2/df
P
interference activity interference, affective interference
0.17 0.05 0.12
0.16e0.18 0.87 35 803.33 0.04e0.07 0.99 20 60.08 0.09e0.14 0.95 30 88.04
3.00 2.94
<0.05 <0.05
interference activity interference, affective interference
0.18 0.07 0.11
0.16e0.21 0.87 35 189.13 0.01e0.10 0.99 22 34.20 0.08e0.14 0.96 29 73.41
1.55 2.53
0.05 <0.05
interference activity interference, affective interference
0.18 0.11 0.12
0.15e0.20 0.86 35 198.34 0.07e0.13 0.96 29 76.03 0.09e0.14 0.95 30 88.04
2.62 2.94
<0.05 <0.05
interference activity interference, affective interference
0.15 0.08 0.07
0.13e0.17 0.86 35 189.75 0.06e0.10 0.96 32 72.93 0.05e0.10 0.97 30 61.10
2.26 2.04
<0.05 <0.05
interference activity interference, affective interference
0.17 0.08 0.09
0.15e0.19 0.82 35 235.78 0.05e0.10 0.97 30 2.17 0.06e0.11 0.96 30 2.47
65.15 74.08
<0.05 <0.05
RMSEA ¼ root mean square error of approximation; CFI ¼ comparative fit index; df ¼ degrees of freedom. a We excluded 36 observations that were multivariate outliers.
pain. For these groups, the construct validity of the modified BPI has not (to our knowledge) been well studied. In addition, our population included persons managed for postoperative pain in the outpatient setting. In a 2006 randomized trial of patients with osteoarthritis,28 the modified short form of BPI exhibited a three-factor structure (pain intensity, activity interference, and affective interference). Yet, recommendations for using the standard short-form BPI suggested a two-factor structure
including pain intensity and activity interference.29 Our findings indicate that a two-factor structure (pain intensity and interference) performed better than did the three-factor model, suggesting the two-factor model may be appropriate for noncancer pain associated with medical conditions physicians see routinely in daily clinical practice. The diversity of the patient cohorts with respect to the underlying cause of pain must be considered. Fear related to pain intensity
Fig. 1. Confirmatory factor model for the two-factor solutiondoverall.
Vol. 47 No. 2 February 2014
Brief Pain Inventory in Non-Cancer Pain Patients
331
Table 4 Standardized Factor Loadings and Factor Correlations by Patient Cohort and Latent Construct for Two-Factor Model Variables Standardized factor coefficient Factor 1: pain intensity Pain at its worst in the last 24 hours Pain on average Pain right now Factor 2: interference Interference with general activity Interference with walking ability Interference with work Interference with mood Interference with relations/other people Interference with sleep Interference with enjoyment of life Factor correlations Factor correlation
Overalla (n ¼ 741)
Postoperative (n ¼ 133)
Arthritis (n ¼ 148)
Back/Neck (n ¼ 202)
Injury/Trauma (n ¼ 204)
0.87 0.88 0.81
0.93 0.95 0.88
0.73 0.89 0.79
0.74 0.86 0.76
0.79 0.84 0.81
0.95 0.66 0.77 0.88 0.71 0.85 0.80
0.91 d 0.84 0.85 0.71 0.85 0.82
0.86 0.81 0.86 0.81 0.72 0.79 0.86
0.80 0.68 0.77 0.77 0.67 0.67 0.79
0.78 0.69 0.80 0.79 0.68 0.66 0.80
0.84
0.88
0.83
0.79
0.68
a
Includes patients with fibromyalgia and neuropathic pain.
may lead to behavior modification.30 This suggests that pain may cause patients to change their activities both directly (because of pain experienced during the activity) and indirectly (because of fear of pain that might be experienced during the activity). Although the motivation for interference with activity was beyond the scope of our study, our findings suggest that both pain intensity and activity interference are important factors to consider when evaluating pain if including both acute (i.e., postoperative pain) and acute exacerbation of chronic pain (i.e., arthritis) populations. Our findings suggest that different aspects of each of the factors were more or less important within each of the subgroup populations we studied, most notable between the postoperative and other chronic pain conditions. Our results concur with previous research11 supporting the notion that BPI activity interference also may be useful to assess medical outcomes. Further assessment of the differential contribution of items within each of the factors may be an important avenue for future research, specifically focusing on elucidating differences between acute and chronic pain. The multigroup analysis we conducted supported the same two-factor configuration, regardless of patient cohort type; however, it did not support metric invariance across the patient cohort type. A path analysis revealed that differences across the patient cohorts were apparent for the three pain measures, general activities, work, and sleep. This is not
completely unexpected, as our patient cohort types included both individuals experiencing acute exacerbation of chronic pain (e.g., neuropathic pain, arthritis, and fibromyalgia) and patient cohort types experiencing acute experiences of pain that is expected to be of relatively short duration (e.g., postoperative pain, injury/trauma). Indeed, an analysis of the means and SDs of pain measures show some variation across groups. With this in mind, perception of pain is likely to vary in situations. For example, a 2004 study by Grotle et al.31 assessed differences in fear avoidance between persons with acute low back pain and a second group with chronic back pain. Lower levels of fear avoidance were observed in the population with acute back pain compared with those with chronic back pain, even after controlling for depression.32 Catastrophizing, attaching negative emotions to pain,32 has been theorized to modify the pain experience. A systematic literature review of catastrophizing behavior in arthritis, fibromyalgia, and other related diseases concluded that catastrophizing is related to pain and pain management-related outcomes.33 These and other studies suggest that relationships between pain assessment and the patient’s pain experience can vary across different acute and chronic conditions that may be associated with pain. The changing regulatory environment has placed increasing importance on the use of patient-reported outcomes in clinical trials,1
332
Lapane et al.
with clear guidance regarding the need for better understanding of the psychometric properties of instruments as used in different patient cohorts.2,3 Additional information regarding the construct validity of the modified BPI in specific patient populations is warranted. BPI is a relevant instrument for use in research for several reasons. First, completing BPI requires little time, and as such, it has been recommended in studies requiring repeated assessment over the course of the study. BPI captures the relevant information needed, with relatively few questions. Furthermore, BPI has been shown to be responsive to change over time in behavioral34 and pharmacological35 interventions. Our findings must be considered with several caveats in mind. The modified BPI includes three measures of pain intensity. As such, the findings of this study may not be generalizable to studies in which BPI with four measures of pain intensity were captured. Although the registry set out to capture at least 200 patients in each of six patient cohorts, recruitment goals were not met in several of the cohorts. Therefore, we did not have a sufficient sample size to evaluate BPI in patients with fibromyalgia and patients with neuropathic pain.
Conclusion Our study was conducted in an outpatient setting with patients experiencing noncancer pain and provides further evidence of the psychometric properties of BPI. We have confirmed the construct validity of BPI in ambulatory patients. Despite the differences in the patient population, our findings reflected the work of others. That is, the confirmatory factor analysis supports the two-factor representation of BPI. As selecting the appropriate instruments is essential for the robustness of any study, this study provides additional support in favor of BPI.
Disclosures and Acknowledgments This work was funded through a consulting agreement with Janssen Scientific Affairs, LLC to Dr. Lapane and Dr. Quilliam. The authors declare no additional potential conflicts of interest.
Vol. 47 No. 2 February 2014
References 1. Bottomley A, Jones D, Claassens L. Patient-reported outcomes: assessment and current perspectives of the guidelines of the Food and Drug Administration and the reflection paper of the European Medicines Agency. Eur J Cancer 2009;45: 347e353. 2. U.S. Food and Drug Administration. Guidance for industry. Patient-reported outcome measures: use in medical development to support labeling claims. 2009. Available from http://www.fda.gov/ downloads/Drugs/GuidanceComplianceRegulatory Information/Guidances/UCM193282.pdf. Accessed March 4, 2012. 3. European Medicines Agency. Committee for Medicinal Products for Human Use (CHMP). Reflection paper on the regulatory guidance for the use of health-related quality of life (HRQL) measures in the evaluation of medicinal products. 2005. Available from http://www.emea.europa.eu/pdfs/human/ ewp/13939104en.pdf. Accessed March 4, 2012. 4. Streiner DL, Norman GR. Health measurement scales. A practical guide to their development and use, 2nd ed. New York: Oxford Medical Publications, Inc., 1995. 5. Atkinson TM, Mendoza TR, Sit L, et al. The Brief Pain Inventory and its ‘‘pain at its worst in the last 24 hours’’ item: clinical trial endpoint considerations. Pain Med 2010;11:337e346. 6. Cleeland CS, Nakamura Y, Mendoza TR, et al. Dimensions of the impact of cancer pain in a four country sample: new information from multidimensional scaling. Pain 1996;67:267e273. 7. Atkinson TM, Rosenfeld BD, Sit L, et al. Using confirmatory factor analysis to evaluate construct validity of the Brief Pain Inventory (BPI). J Pain Symptom Manage 2011;41:558e565. 8. Atkinson TM, Halabi S, Bennett AV, et al. Cancer and Leukemia Group B. Measurement of affective and activity pain interference using the Brief Pain Inventory (BPI): Cancer and Leukemia Group B 70903. Pain Med 2012;13:1417e1424. 9. Brown TA. Confirmatory factor analysis for applied research. New York: Guilford, 2006. 10. Kim M, Chow W, Benson C. Rationale and design of the Oxycodone Users Registry: A prospective, multicenter registry of patients with nonmalignant pain. J Opioid Manag 2013;9(3): 189e204. 11. Keller S, Bann CM, Dodd SL, et al. Validity of the Brief Pain Inventory for use in documenting the outcomes of patients with noncancer pain. Clin J Pain 2004;20:309e318. 12. Farrar JT, Young JP Jr, LaMoreaux L, Werth JL, Poole RM. Clinical importance of changes in chronic pain intensity measured on an 11-point numerical pain rating scale. Pain 2001;94:149e158.
Vol. 47 No. 2 February 2014
Brief Pain Inventory in Non-Cancer Pain Patients
13. Farrar JT, Berlin JA, Strom BL. Clinically important changes in acute pain outcome measures: a validation study. J Pain Symptom Manage 2003;25: 406e411. 14. Farrar JT, Portenoy RK, Berlin JA, Kinman JL, Strom BL. Defining the clinically important difference in pain outcome measures. Pain 2000;88: 287e294. 15. Schreiber JB, Stage FK, King J, Nora A, Barlow EA. Reporting structural equation modeling and confirmatory factor analysis results: a review. J Educ Res 2006;99:323e337. 16. Bollen KA. Structural equations with latent variables, 2nd ed. New York: Wiley, 1995. 17. Kline RB. Principles and practice of structural equation modeling, 2nd ed. New York: Guilford, 2005. 18. Browne MW, Cudeck R. Alternative ways of assessing model fit. Newbury Park, CA: Sage, 1993. 19. Bentler PM. Comparative fit indexes in structural models. Psychol Bull 1990;107:238e246. 20. Hu L, Bentler PM. Cutoff criteria for fit indexes in covariance structure analysis: conventional criteria versus new alternatives. Struct Equation Model 1999;6:1e55. 21. Daut RL, Cleeland CS. The prevalence and severity of pain in cancer. Cancer 1982;50:1913e1918. 22. Amos [computer program], version 20. Meadville, PA: AMOS Develpoment Corporation, 2012. 23. Breitbart W, McDonald MV, Rosenfeld B, et al. Pain in ambulatory AIDS patients. I: Pain characteristics and medical correlates. Pain 1996;68:315e321. 24. Beauregard L, Pomp A, Choiniere M. Severity and impact of pain after day surgery. Can J Anaesth 1998;45:304e311. 25. Zalon ML. Comparison of pain measures in surgical patients. J Nurs Meas 1999;7:135e152. 26. Roth SH, Fleischmann RM, Burch FX, et al. Around-the-clock, controlled-release oxycodone
333
therapy for osteoarthritis-related pain: placebocontrolled trial and long-term evaluation. Arch Intern Med 2000;160:853e860. 27. Semenchuk MR, Sherman S, Davis B. Doubleblind, randomized trial of bupropion SR for the treatment of neuropathic pain. Neurology 2001; 57:1583e1588. 28. Mendoza T, Mayne T, Rublee D, Cleeland C. Reliability and validity of a modified Brief Pain Inventory short form in patients with osteoarthritis. Eur J Pain 2005;10:353e361. 29. Turk DC, Dworkin RH, Allen RR, et al. Core outcome domains for chronic pain clinical trials: IMMPACT recommendations. Pain 2003;106: 337e345. 30. Vlaeyen JWS, Kole-Snijders AMJ, Boren RGB, VanEek H. Fear of movement/(re)injury in chronic low back pain and its relation to behavioral performance. Pain 1995;62:363e372. 31. Grotle M, Vollestad NK, Veierod MB, Brox JI. Fear-avoidance beliefs and distress in relation to disability in acute and chronic low back pain. Pain 2004;112:343e352. 32. Sullivan MJ, Thorn B, Haythornthwaite JA, et al. Theoretical perspectives on the relation between catastrophizing and pain. Clin J Pain 2001;17: 52e64. 33. Edwards RR, Bingham CO, Bathon J, Haythornthwaite JA. Catastrophizing and pain in arthritis, fibromyalgia, and other rheumatic diseases. Arthritis Rheum 2006;55:325e332. 34. Twycross R, Harcourt J, Bergl S. A survey of pain in patients with advanced cancer. J Pain Symptom Manage 1996;12:273e282. 35. Hwang SS, Chang VT, Kasimis B. Dynamic cancer pain management outcomes: the relationship between pain severity, pain relief, functional interference, satisfaction and global quality of life over time. J Pain Symptom Manage 2002;23:190e200.