- Email: [email protected]
Validation of the Modified Brief Pain Inventory-Exploratory Form in Surgery Patients
ORIGINAL RESEARCH ARTICLE
Validation of the Modified Brief Pain Inventory-Exploratory Form in Surgery Patients Wen-Hung Chen, PhDa, Kitty S. Chan, PhDb, Tong J. Gan, MDc, Connie Chen, PharmDd, Mani Lakshminarayanan, PhDd,1, Dennis A. Revicki, PhDa a
CENTER FOR HEALTH OUTCOMES RESEARCH, UNITED BIOSOURCE CORPORATION, BETHESDA, MARYLAND; DEPARTMENT OF HEALTH POLICY AND MANAGEMENT, JOHNS HOPKINS BLOOMBERG SCHOOL OF PUBLIC HEALTH, BALTIMORE, MARYLAND; cDEPARTMENT OF ANESTHESIOLOGY, DUKE UNIVERSITY MEDICAL CENTER, DURHAM, NORTH CAROLINA; AND dPFIZER, NEW YORK, NEW YORK
b
-
ABSTRACT
O B J E C T I V E : An exploratory version of the Modified Brief Pain Inventory (mBPI-e) to measure acute post-operative pain, with new items on coughing, breathing, and concentration, was examined for their measurement properties. S T U D Y D E S I G N : This is a secondary study using data from two randomized clinical trials: general surgery trial (N=1050) and coronary artery bypass graft (CABG) surgery trial (N=1636). The measurements used in the two trials were: 1) mBPI-e; 2) clinician and patient global evaluations of medications; and 3) pain intensity diary. The mBPI-e and pain intensity were collected for 10 days. Clinician and patient global evaluations of medication were collected twice. The analyses conducted were: 1) exploratory factor analysis (EFA); 2) confirmatory factor analysis (CFA); 3) item response theory (IRT); 4) internal consistency; 5) test-retest reliability; 6) concurrent validity; 7) known-group validity; and 8) responsiveness. R E S U L T S : Pain severity, pain interference, and coughing and breathing factors were identified. Pain severity and pain interference subscale scores were constructed for mBPI-e. IRT analyses showed all items exhibited good item characteristics. Internal consistency was 0.85 for severity and 0.87 for interference. Test-retest reliability was 0.81 for severity and 0.71 for interference. Both severity and interference scores were correlated with diary-based pain intensity ratings (P <.0001). Mean severity and interference scores varied by physician and patient global ratings (P <.05). Severity and interference scores were responsive to changes in pain diary scores and physician global ratings (P <.001). There were no substantive differences in reliability or validity for sub-samples of surgery patients. C O N C L U S I O N S : The original BPI has been used in clinical studies, and the mBPI has demonstrated good reliability and validity in CABG patients. Based on this study, the mBPI-e has also demonstrated good reliability and validity for assessing postoperative acute pain in CABG and general surgery patients. See funding, conflict of interest, and authorship disclosures at the end of this article. Ó 2010 Elsevier Inc. All rights reserved Health Outcomes Research in Medicine (2010) 1, e17-e28
e18
Modified Brief Pain Inventory-Exploratory Form
K E Y W O R D S : Brief pain inventory; Measurement of pain; Modified brief pain inventory; Pain measure; Post operative pain
Postoperative
pain is common after highly invasive procedures such as coronary artery bypass graft (CABG) surgery, abdominal surgery, major orthopedic surgery, and gynecologic surgery. Clinical trials comparing different analgesic medications require psychometrically sound assessments of pain severity and interference with everyday activities. The Brief Pain Inventory (BPI) has been used in numerous studies and clinical trials of chronic and acute pain.1-7 In the postsurgical setting, ability to engage in normal daily activities is limited, and general pain assessments such as the pain interference scale from the BPI2 may not be entirely applicable. Mendoza et al8 developed a modified version of the BPI (mBPI) in CABG surgery patients, where items on general activity, normal work, and enjoyment of life were removed from the BPI and a new item on concentration was added. They found good evidence for internal consistency and test-retest reliability, confirmed the 2-domain factor structure and evidence supporting concurrent validity of the mBPI.
The objective of the current study was to evaluate the psychometric characteristics (ie, reliability, validity, responsiveness) of an exploratory modified version of the mBPI using data collected in two postoperative pain clinical trials. This study is the secondary data analysis of these 2 safety and efficacy trials. The exploratory version of the mBPI (mBPI-e) included 2 new items on interference related to coughing and deep breathing, which were intended to make the mBPI more relevant for assessing pain outcomes in postoperative analgesic clinical trials. The mBPI-e is based on the mBPI,8 and retains the original 4 pain severity items from the original BPI (worst, least, and average pain in the past 24 hours, and pain right now). For the new pain F I G U R E 1 : Items included in the BPI, mBPI, and mBPI-e. interference scale, mBPI-e retains the original general activity item and added 2 new items on coughing and deep breathing. These new items were added aiming to enhance the instrument’s ability in assessing acute pain outcomes in postoperative clinical studies. The assumption was that the interference associated with coughing and deep breathing can be very significant in many postoperative patients, depending on the type of surgery (eg, CABG, gastrointestinal [GI]). Figure 1 compares the items contained in BPI, mBPI, and mBPI-e.
METHODS Data for this study were taken from 2 clinical studies with different inclusion/exclusion criteria. Study 069 was a randomized, double-blind, parallel-group, multicenter study of the safety and efficacy of parecoxib followed by valdecoxib compared with placebo in general surgery patients for treatment of postoperative pain. It was conducted in 113 centers in 14 countries from September 2002 to February 2003.9 Study 071 was a randomized, double-blind, parallel-group, multicenter study of the safety and efficacy of parecoxib/valdecoxib and placebo/valdecoxib compared with placebo for treatment of postoperative pain in patients who had CABG. It was conducted at 175 centers in 27 countries from January 2003 to January 2004.10 Both Phase
Health Outcomes Research in Medicine - Vol. 1 / No. 1 / July 2010
III clinical trials were conducted with 10 days of treatment and 30 days of follow-up after study drug administration was completed.
Measures Modified BPI-e: The mBPI-e was administered each night at bedtime, with a 24-hour recall starting the day after surgery to examine pain severity (worst, least, average, and current pain) and pain interference (general activity, mood, walking ability, relations with others, sleeping, coughing, deep breathing, and concentration). Responses for these mBPI-e items were captured through the use of 11epoint numeric rating scales anchored at 0 (no pain or does not interfere) and ranged through 10 (pain as bad as you can imagine or completely interferes). The mBPI-e yielded 2 scale scores: pain severity and pain interference. The 2 scale scores were computed as the means of their corresponding item ratings. The higher scores indicate greater pain intensity or increased functional impairment. Imputation of missing item responses was based on the 50% rule, that is, if there were more than 50% missing item responses, the score for that scale was set to missing. When < 50% of item responses were missing, we used the average of the nonmissing item responses to impute missing item responses. The scale score then was calculated as the mean of the nonmissing item responses. There were very few missing item response data, and no more than 0.24% of the total sample required imputed pain severity or interference scale scores.
Global Evaluation of Treatment: Patient and physician global evaluations of study medication were collected through the use of a 4-point scale (1 ¼ poor, 2 ¼ fair, 3 ¼ good, 4 ¼ excellent). Physicians recorded their global evaluations when they examined the patients at the time of patient transition from parenteral to oral study medication (IV/PO), and at the end of the study (or early termination). The patients also were asked to record their own global evaluations at the time of IV/PO transition, and at the end of study (or early termination) when they returned to the clinical site for a final visit.
Daily Pain Diary: Patients recorded their current pain level in a diary through the use of a verbal rating scale (0 ¼ none, 1 ¼ mild, 2 ¼ moderate, 3 ¼ severe) each day from baseline (Day 1) through Day 9 prior to the AM dosing (usually at 8:00 AM) and at 2 (ie, 10:00 AM), 4 (12:00 PM), 8 (4:00 PM), and 12 (8:00 PM) hours (prior to PM dosing) and at 24 hours (8:00 AM the next day, prior to AM dosing). Average pain intensity (API) was calculated by using the trapezoid rule to obtain the sum across the hours, then average over the hours between the first and the last valid ratings, and had a range of 0 to 3.
Analysis Populations The psychometric analyses used all intention-to-treat patients who received at lease 1 dose of the study medication. The pooled general and CABG surgery data were structured by numbered study days. For consistency with the effect of the medication to the level of pain, this study designated baseline measurement as the day after the surgery.
Psychometric Analyses One focus of this evaluation study was the inclusion of the 2 new items (ie, coughing and deep breathing) in the pain interference scale. Therefore, we examined whether these 2 new items measure the same construct as the other original pain interference items, and whether they add additional validity to the instrument. We also assessed the item performance separately for the CABG and GI subgroups, and the nonGI subgroup, with focus on the coughing and deep breathing items. This was because these 2 items were more specific to the CABG and GI patients, and the item characteristics might be different between the 2 subgroups.
Item Analysis: We assessed mBPI-e item performance using data at baseline. We conducted exploratory factor analysis and confirmatory factor analysis to assess the dimensionality. We conducted item response
e19
e20
Modified Brief Pain Inventory-Exploratory Form
theory (IRT) analysis to assess whether the item response categories were monotonically ordered (ie, from lowest response option to highest response option) for each item and whether each item associated strongly with the underlying construct (ie, pain). For the factor analysis and IRT analysis, the sample was randomly divided into 2 subgroups. Exploratory factor analysis was conducted using 1 subgroup, and the confirmatory factor analysis was performed using the other subgroup. Promax rotation was applied to the factor structure, allowing the derived factors to be correlated. Eigenvalue > 1 was used as the criterion to extract factors. For the IRT analyses we used the Samejima graded response model11 and MULTILOG.12
Reliability: After the analysis of the item characteristics and item performance, we determined the final content of the mBPI-e. We then assessed the internal consistency reliability (Cronbach alpha) using baseline data. We assessed the test-retest reliability using data from Day 5 to Day 9 with a stable group of patients. Stable patients were defined as no change in their API scores within these assessment intervals. Intraclass correlation coefficients (ICCs) based on analysis of variance models were used to assess test-retest reliability.13
Construct Validity: We assessed the convergent validity by correlating the mBPI-e scale scores with the average rating of pain intensity for baseline, Day 3, Day 5, and Day 9. We assessed the known group validity of the instrument by comparing the mean mBPI-e scale scores across the classifications by the physician’s global evaluation of study medication and by the patient’s global evaluation of study medication.
Responsiveness: We assessed the responsiveness of the instrument by comparing the score change between responders and nonresponders from baseline to Day 3, and from baseline to Day 9 in the overall sample. Two definitions of responders and nonresponders were created, one based on the API, and the other by the physician global evaluation of study medication. Using the API, we defined the responder as patients with at least 1 point decrease in API scores from baseline to the target study day. This also corresponds to at least 1 level lower in pain intensity. A half-point F I G U R E 2 : Days of data used for psychometric decrease in API was defined as analyses. a borderline responder, and all other patients were defined as nonresponders. Analysis of covariance was conducted, testing mean score changes among the 3 groups, controlling for gender, age, and mBPI-e score at baseline. Based on the physician’s global evaluation, responders were defined as those patients who had good or excellent ratings, and all other patients were considered nonresponders. Analysis of covariance models, adjusting for gender, age, and mBPI-e score at baseline, were conducted to compare mean score differences for responders and nonresponders. Finally, the effect size by responders and nonresponders based on the API was estimated. Effect size was calculated as the difference in raw mean scores for responders and for nonresponders from baseline to endpoint (Day 3 or Day 9) divided by the standard deviation of the scale score at baseline.14 In all the analyses, nominal P-values were reported and no adjustments were made for any multiplicity. Statistical significance was claimed when P-value was <.05. Figure 2 shows the days of the data used for each of the analyses above.
Health Outcomes Research in Medicine - Vol. 1 / No. 1 / July 2010
e21
RESULTS Patient Demographics In the combined sample, mean age was 58.5 years (standard deviation [SD] ¼ 12.1) and the group was mostly male (68.4%) and white (93.2%) (Table 1). There were some differences between the general surgery and the CABG studies, with older patients (mean 61.9 vs 53.3 years) and more males in the CABG study.
Item Analyses Item Characteristics and Correlations: For the pain intensity and interference items, the entire range of item responses was observed. For pain intensity items, there was no evidence of floor or ceiling effects. For pain interference items, there was no evidence of a ceiling effect (range 1%-10%), but some evidence of a floor effect (range 11%-33%). At baseline, for pain interference, the highest reported interference was on general activity (mean 4.6, SD 3.3) and walking ability (mean 4.6, SD 3.8), and the least reported interference was on relations with others (mean 2.0, SD 2.7) and concentration (mean 2.2, SD 2.8). The coughing and breathing items were more likely to be endorsed by the GI and CABG surgery group (coughing: mean 3.9, SD 3.4; breathing: mean 4.0, SD 3.2) compared with the non-GI surgery group (coughing: mean 2.5, SD 3.3; breathing: mean 1.7, SD 2.6). The inter-item correlations ranged from 0.41 (worst and least pain) to 0.73 (worst and average pain) within pain severity, and ranged from 0.24 (sleep and coughing) to 0.67 (general activity and walking ability) within pain interference. The correlations between the pain severity and interference items ranged from 0.10 (least pain and coughing) to 0.52 (worst pain and general activity).
Factor Analysis: The exploratory factor analysis was conduced using one random split-half sample. Promax rotation was applied to the factor structure to allow the factors to be correlated. The results supported a 3factor solution with eigenvalues of 5.66, 1.57, and 1.10. Further inspection of the factor loading found that - T A B L E 1 : Demographics for Modified Intent-to-Treat Population* (n ¼ 2686) from General Surgery Study 069 (n ¼ 1050) and CABG Study 071 (n ¼ 1636)
Age (years), mean (SD) Gender, n (%) Female Male Race, n (%) White Black Asian Not listed Surgery type, n (%) Orthopedic Gastrointestinal Gynecologic Thoracic CABG Other surgery not listed
Combined
General Surgery Study
58.5 (12.13)
53.3 (14.35)
848 (31.6%) 1838 (68.4%)
617 (58.8%) 433 (41.2%)
231 (14.1%) 1405 (85.9%)
2504 (93.2%) 77 (2.9%) 58 (2.2%) 47 (1.7%)
977 (93.0%) 47 (4.5%) 3 (0.3%) 23 (2.2%)
1527 (93.3%) 30 (1.8%) 55 (3.4%) 24 (1.5%)
279 (10.4%) 391 (14.6%) 204 (7.6%) 22 (0.8%) 1636 (60.9%) 199 (7.4%)
279 (26.6%) 391 (37.2%) 204 (19.4%) 22 (2.1%) 0 (0.0%) 199 (19.0%)
0 (0.0%) 0 (0.0%) 0 (0.0%) 0 (0.0%) 1636 (100.0%) 0 (0.0%)
CABG ¼ coronary artery bypass graft; SD ¼ standard deviation. *Patients randomized to treatment who received at least 1 dose of study medication.
CABG Study 61.9 (8.96)
e22
Modified Brief Pain Inventory-Exploratory Form
factors could be identified as pain severity, pain interference, and a third factor consisting of the breathing and coughing items (Table 2). Given that the coughing and breathing items are more relevant for the GI and CABG patients, the factor analysis was repeated, including only the GI and CABG surgery patients. In this factor analysis, only 2 factors were extracted (eigenvalues: 6.24 and 1.42): pain severity and pain interference, and the coughing and breathing items were in the pain interference factor. The confirmatory factor analysis was conducted using the other split-half sample with 2-factor structure: 4 items for pain severity and 8 items for pain interference. Bentler’s comparative fit index (CFI) was 0.859, the standardized root mean residual (SRMR) was 0.067, and the root mean square error of approximation (RMSEA) was 0.129 (90% confidence interval, 0.122-0.137). These statistics did not meet the acceptable good-fit criteria (>0.90 for CFI, and <.05 for SRMR and RMSEA). The confirmatory factor analysis also was conducted using the 3-factor structure: 4 items for pain severity, 6 items for pain interference, and 2 coughing and breathing items. CFI was 0.910, the SRMR was 0.050, and the RMSEA was 0.105 (90% confidence interval, 0.097-0.113). A good fit to the observed data was accepted based on CFI and SRMR having met the criteria, even though RMSEA had not. We believe that all 3 fit statistics would reach the acceptable range if we 1) increase the number of factors; or 2) reduce the number of items. On the other hand, evidence of eigenvalues and factor loadings (Table 2) discouraged increase in the number of factors. And the content coverage (in the light of content validity) discouraged deleting the items. Therefore, we decided to accept the factor structure and acknowledge that the RMSEA did not meet the good-fit criterion.
Item Response Theory Analysis: We conducted an IRT analysis to assess the psychometric properties of the items. For pain severity, the 4 items covered the range of the pain severity domain. However, the average pain item has a substantially large slope parameter (7.58), which might be due to redundancy with the other 3 pain severity items (2.01, 2.02, 2.18). For pain interference, slope parameters ranged from 0.90 (for coughing) to 2.88 (for mood). Separate IRT analysis for the other split-half sample yielded similar findings.
Item Review Summary: Based on the item analyses, factor analyses, and the IRT analyses, the mBPI-e clearly contains 2 domains: pain severity and pain interference. Based on the IRT analyses of the pain severity items, fewer items are sufficient to assess pain severity or intensity. However, from the content and face validity perspective and for consistency with the original BPI, all 4 items were retained.
- T A B L E 2 : Rotated Factor Pattern* of mBPI-e from Exploratory Factor Analysis Combined Group with Split-Half Sample (n ¼ 1340)†
Relations with others (Interference) Mood (Interference) Concentration (Interference) Sleep (Interference) Walking ability (Interference) General activity (Interference) Average (Severity) Least pain (Severity) Pain right now (Severity) Worst pain (Severity) Coughing (Interference) Breathing (Interference)
Factor 1
Factor 2
Factor 3
0.854 0.833 0.738 0.734 0.732 0.715 0.038 0.110 0.081 0.142 0.084 0.120
0.062 0.062 0.088 0.058 0.045 0.154 0.910 0.878 0.790 0.679 0.013 0.010
0.009 0.038 0.198 0.040 0.053 0.034 0.009 0.015 0.004 0.043 0.950 0.848
mBPI-e ¼ Modified Brief Pain Inventory-exploratory form. *Standardized Regression Coefficients after PROMAX oblique rotation. † Of 1340 observations, 525 are omitted due to missing values; 815 actually used in analysis.
Health Outcomes Research in Medicine - Vol. 1 / No. 1 / July 2010
For the pain interference scale, the results show that the item on concentration does measure the same construct as with other pain inference items. The coughing and breathing items were more relevant for the GI and CABG patients than with other general surgery patients. These items do not consistently separate into a third factor, and it was uncertain whether these new items add to the measurement of pain interference in all patients. Therefore, we conducted the following psychometric analyses based on 3 sets of data: 1) the combined sample was analyzed with and without the coughing and breathing items; 2) GI and CABG patient samples were analyzed with and without the coughing and breathing items; and 3) the other general surgery patient subset was analyzed, with and without the coughing and breathing items. The intent was to compare the reliability and validity by different subgroups with and without these 2 items to evaluate their contribution to assessing pain interference.
Reliability Internal Consistency Reliability, and Test-retest Reliability: The internal consistency reliability, at baseline, was 0.85 for the pain severity scale and 0.87 for the pain interference scale (with or without the coughing and breathing items). For the GI and CABG patients, internal consistency was 0.89 with the coughing and breathing items and 0.88 without these items. For the other general surgery patients, internal consistency reliability was 0.83 and 0.85 with and without the coughing and breathing items, respectively. Test-retest reliability for pain severity was 0.81. For pain interference, the ICCs were 0.75 and 0.73 for the overall sample with and without the coughing and breathing items, respectively. For the GI and CABG patients, ICCs for pain interference were 0.74 with the coughing and breathing items and 0.71 without the 2 items. For the general surgery group, ICCs were 0.81 and 0.78 with and without the 2 items, respectively.
Validity Construct Validity: Construct validity for mBPI-e was assessed by the correlations between the pain severity and interference scale scores and the API for Days 1, 3, 5, and 9. As expected, the correlation between the pain severity and API is higher than the correlation between the pain inference and API. The correlations between the pain severity scale score and the API ranged from 0.57 to 0.67 over these assessments (all P <.001). The correlations between the pain interference scale score and the API ranged from 0.39 to 0.49 (all P <.001). The correlations between pain interference and API with or without the coughing and breathing items were not substantively different (data not shown). Known-group Validity: Known-group validity assesses the ability of the mBPI-e scale scores to distinguish between different groups of patients. For this study, we based the examination of known-group validity on the physician’s global evaluation of study medication (Table 3) and the patient’s global evaluation of study medication (Table 4). Based on the physician’s global evaluation, at IV/PO switch, both mBPI-e scale scores were statistically significantly different between the 4 groups (all P <.001), except between poor and fair categories (P >.05). At end of study, the mBPI-e scale scores were significantly different (all P <.05), except between good and excellent groups (P >.05). Based on the patient’s global evaluation, the mBPI-e scale scores were significantly different between every pair of groups at IV/PO switch (all P <.001). At end of study, the mBPI-e scale scores were significantly different (all P <.05), except those between fair and good groups (P >.05), and between good and excellent groups for pain interference (P >.05). Overall, both mBPI-e scale scores show very good known-group validity in terms of mean score differences based on the physician global evaluations of study medication. These findings were not substantively different with or without the coughing or breathing items, or by surgical subgroup (data not shown).
Responsiveness: API-defined responders reported significantly greater reductions in pain severity, with mean decreases of 1.93 points, compared to mean decreases of 0.80 point for nonresponders (P <.0001) and
e23
e24
Modified Brief Pain Inventory-Exploratory Form
- T A B L E 3 : Mean mBPI-e Scores by Physician Global Evaluation of Study Medication at IV/PO Switching and End of Study Modified Intent to Treat Population (n ¼ 2686) Physician Global Evaluation of Study Medication mBPI-e Scales At IV/PO switch Pain severity Pain interference At end/early termination of study Pain severity Pain interference
Poor n, Mean (SD)
Fair n, Mean (SD)
Good n, Mean (SD)
Excellent n, Mean (SD)
P-Value*
122, 3.7 (1.92) 122, 3.2 (2.06)
281, 3.3 (1.77) 284, 2.7 (2.00)
600, 2.7 (1.66) 603, 2.1 (1.82)
328, 2.0 (1.41) 327, 1.6 (1.60)
2§,3§,4§,5§,6§ 2§,3§,4‡,5§,6‡
72, 3.6 (2.56) 72, 2.8 (2.79)
99, 2.7 (1.90) 99, 1.9 (2.05)
204, 1.9 (1.64) 205, 1.3 (1.51)
127, 1.4 (1.41) 127, 0.9 (1.11)
1†,2§,3§,4†,5§ 1†,2§,3§,4†,5‡
IV ¼ intravenous (parenteral); mBPI-e ¼ Modified Brief Pain Inventory-exploratory form; PO ¼ oral; SD ¼ standard deviation. *One-way analysis of variance using Scheffe’s post hoc comparisons for differences between groups; 1 ¼ poor vs fair; 2 ¼ poor vs good; 3 ¼ poor vs excellent; 4 ¼ fair vs good; 5 ¼ fair vs excellent; 6 ¼ good vs excellent. † P <.05. ‡ P <.001. § P <.0001.
compared with borderline responders (1.93 vs 1.38, P <.05) at Day 3 (Table 5). Comparable significant differences were observed in mean changes in pain severity scores between responders and nonresponders (2.90 vs 1.39, P <.001) and responders and borderline responders (2.90 vs 2.28, P <.001) at Day 9. Responders reported significantly more reductions in pain interference compared with nonresponders (1.84 vs 1.02, P <.05) on Day 3. By Day 9, API-based responders reported significant differences on pain inference compared with borderline responders (3.03 vs 2.63, P <.05) and with nonresponders (3.03 vs 1.87, P <.0001). At IV/PO switch, physician global rating-defined responders reported mean decreases of 1.23 points in pain severity scores compared with mean decreases of 0.69 points in nonresponders (P <.0001) (Table 6). At end of study, responders reported significantly greater decreases in pain severity scores compared with nonresponders (2.22 vs 1.33, respectively, P <.0001). Also, responders reported significantly greater
- T A B L E 4 : Mean of mBPI-e Scores by Patient Global Evaluation of Study Medication at IV/PO Switching and End of Study Modified Intent to Treat Population (n ¼ 2686) Physician Global Evaluation of Study Medication mBPI-e Scales At IV/PO switching Pain severity Pain interference At end/early termination of study Pain severity Pain interference
Poor n, Mean (SD)
Fair n, Mean (SD)
Good n, Mean (SD)
Excellent n, Mean (SD)
P-Value*
93, 4.2 (1.92) 93, 3.7 (2.19)
258, 3.2 (1.61) 260, 2.6 (1.99)
651, 2.7 (1.69) 655, 2.1 (1.76)
328, 2.1 (1.58) 326, 1.6 (1.65)
1‡,2§,3§,4‡,5§,6§ 1§,2§,3§,4‡,5§,6‡
72, 4.2 (2.49) 72, 3.4 (2.97)
81, 2.5 (1.65) 81, 1.7 (1.44)
227, 1.9 (1.63) 227, 1.2 (1.43)
120, 1.3 (1.33) 121, 0.7 (1.02)
1§,2§,3§,5§,6† 1§,2§,3§,5‡
IV ¼ intravenous (parenteral); mBPI-e ¼ Modified Brief Pain Inventory-exploratory form; PO ¼ oral; SD ¼ standard deviation. *One-way analysis of variance using Scheffe’s post hoc comparisons for differences between groups; 1 ¼ poor vs fair; 2 ¼ poor vs good; 3 ¼ poor vs excellent; 4 ¼ fair vs good; 5 ¼ fair vs excellent; 6 ¼ good vs excellent. † P <.05. ‡ P <.001. § P <.0001.
Health Outcomes Research in Medicine - Vol. 1 / No. 1 / July 2010
e25
- T A B L E 5 : The Responsiveness of mBPI-e by Average Pain Intensity (Responder vs Nonresponder) mITT Population (n ¼ 2686) Classification by change in Average Pain Intensity* ‡
mBPI-e Scales From baseline to day 3 Pain severity Pain interference From baseline to day 9 Pain severity Pain interference
Responder n, LS Mean (SE)
Borderline Responder§ n, LS Mean (SE)
Nonresponderk n, LS Mean (SE)
Overall Model† P-Value
Pair Comparison{
137, 1.93 (0.125) 137, 1.84 (0.143)
280, 1.38 (0.087) 280, 1.45 (0.100)
832, 0.80 (0.052) 831, 1.02 (0.060)
<.0001 <.0001
1††,2††, 3# 1#, 2#
152, 2.90 (0.118) 152, 3.03 (0.107)
148, 2.28 (0.120) 148, 2.63 (0.110)
332, 1.39 (0.083) 334, 1.87 (0.076)
<.0001 <.0001
1††,2††, 3** 1††,2††, 3#
LS ¼ least square; mBPI-e ¼ Modified Brief Pain Inventory-exploratory form; mITT ¼ Modified Intent to Treat; SE ¼ standard error. *Average Pain Intensity (API) is computed using trapezoid rule on observed ratings such that API ¼ 0.5 {2 (PI0 þ PI2) þ 2 (PI2 þ PI4) þ 4 (PI4 þ PI8) þ 4 (PI8 þ PI12) þ 12 (PI12 þ PI24)}. † Analysis of covariance controlling for age, gender, and baseline mBPI-e scale score. ‡ API decreases more than 1 point. § API decreases more than 0.5 point but less than 1 point. k API decreases less than 0.5 point. { 1 ¼ nonresponder vs borderline responder; 2 ¼ nonresponder vs responder; 3 ¼ borderline responder vs responder. # P <.05. **P <.001. †† P <.0001.
reductions in pain interference compared with nonresponders (P <.05) at IV/PO switch. By end of study, responders reported mean decreases of 2.43 points compared with mean decreases of 1.75 points in nonresponders (P <.0001). There was no difference in the responsiveness of the interference scale with or without the coughing and breathing items.
- T A B L E 6 : The Responsiveness of mBPI-e by Physician Global Evaluation of Study Medication (Responder vs Nonresponder), mITT Population (n ¼ 2686) Physician Global Evaluation of Study Medication
mBPI-e Scales At IV/PO switching Pain severity Pain interference At end/early termination of study Pain severity Pain interference
Responder† n, LS Mean (SE)
Nonresponder‡ n, LS Mean (SE)
Overall Model P-Value*
Pair Comparison§
835, 1.23 (0.050) 836, 1.36 (0.056)
379, 0.69 (0.076) 380, 0.98 (0.083)
<.0001 <.0001
1{ 1k
309, 2.22 (0.098) 311, 2.43 (0.093)
167, 1.33 (0.135) 165, 1.75 (0.129)
<.0001 <.0001
1{ 1{
IV ¼ intravenous (parenteral); LS ¼ least square; mBPI-e ¼ Modified Brief Pain Inventory-exploratory form; mITT ¼ Modified Intent to Treat; PO ¼ oral; SE ¼ standard error. *Analysis of covariance controlling for age, gender, and baseline mBPI-e scale score. † Global evaluation Good or Excellent. ‡ Global evaluation Poor or Fair. § Responder vs Nonresponder. k P <.001. { P <.0001.
e26
Modified Brief Pain Inventory-Exploratory Form
Effect Size: For pain severity response based on the API at Day 3, the effect size for responders was 1.25 and for nonresponders was 0.37. The effect sizes for pain interference at Day 3 were lower, but different among the 3 responder groups. Effect sizes for pain interference were 0.93, 0.68, and 0.39 for the responder, borderline responder, and nonresponder groups, respectively. At Day 9, the responder group’s effect size for pain severity was 1.81, for borderline responders it was 1.33, and for nonresponders it was 0.64. The effect sizes for pain interference at Day 9 were 1.52, 1.22, and 0.70 for the responder, borderline responder, and nonresponder groups, respectively.
DISCUSSION The data used in this study for the psychometric analysis of mBPI-e combined the clinical trial data that were reported in both Nussmeier et al9 and Ott et al,10 who studied the safety and analgesic efficacy of parecoxib and valdecoxib in comparison with placebo. The mBPI-e data were not reported in Nussmeier et al (2006; study 069),9 and only group comparison results for individual items were reported in Ott et al (2003; study 071).10 This study evaluated the item characteristics and psychometric qualities of the mBPI-e based on a secondary analysis of clinical trial data. The goal was to evaluate the reliability, validity, and responsiveness of the mBPI-e for assessing acute pain in a postoperative setting. Our study found that the mBPI-e is a reliable, valid, and responsive measure of postoperative pain, which could be especially useful for evaluating the effectiveness of the pain treatment for patients who have undergone surgery. Based on the item analyses, no problems were identified for any items that warranted deletion from the revised pain measure. The exploratory factor analyses supported a 3-factor (domain) structure when general, GI, and CABG surgery types of data were included in the analyses, with the new breathing and coughing items forming the third factor. This was likely attributable to the fact that the breathing and coughing items were not as relevant for the general surgery patients. When including only the GI and CABG surgery data, the exploratory and confirmatory factor analyses supported the 2-factor solution with pain severity and pain interference. In the GI and CABG patients, the breathing and coughing items contribute to the pain interference factor. The pain severity and pain interference scales both showed acceptable internal consistency and testretest reliability. Therefore, reliability for the pain severity and interference scales also was supported in this study. The pain severity and interference scales showed evidence supporting construct validity and discriminant validity in this sample of CABG and general surgery patients. As expected, pain severity and API diary scores were moderately and significantly correlated. There were significant correlations between the diarybased pain measures and pain interference, indicating that greater pain severity was associated with more interference in activities, walking, mood, and other aspects of function. Average pain severity and pain interference scores also varied significantly by patient-rated and clinician-rated global evaluations of treatment. For example, the mean pain severity scores for the excellent or good groups based on physician ratings were significantly better than those rated as poor at the end of the study (mean score 1.4 and 1.9 vs 3.6). Similar differences between physician-defined groups were observed for the pain interference scores as well, with the mean scores in the excellent group 0.9 points lower (better) than the good group, 1.5 points lower than the fair group, and 1.9 points lower than the poor group. The mean differences based on patient global ratings were comparable and supported the differences observed based on clinician ratings. Responsiveness was demonstrated for both the pain severity and pain interference scales related to changes in pain severity based on daily diaries and by physician global evaluations of treatment effects. By the end of the study, pain diary responders decreased their pain severity scores by 2.9 points, compared with decreases of 1.4 points in nonresponders. For pain interference, responders reported 3.0 point decreases (improvements), whereas nonresponders reported 1.9 point decreases. For clinical trials, evidence of responsiveness is important, and based on this study, the mBPI-e pain severity and interference scales are highly responsive to changes in daily pain ratings and physician global evaluations.
Health Outcomes Research in Medicine - Vol. 1 / No. 1 / July 2010
The mBPI-e contains 2 new items that extend the coverage of concepts for pain studies in the postoperative hospital setting. The new items on coughing and breathing were more relevant for GI and CABG patients than for other general surgery patients. However, extended and intensive analyses on reliability and validity show that the 2 items can be included or excluded in the instrument and do not effect the measurement properties of the pain interference scale, including responsiveness. Therefore, based on these findings, the recommendation was to include these 2 items in the interference scale when assessing GI and CABG patients, because they add content validity and cover additional areas relevant for GI and CABG surgery patients. The concentration item fits in with the other original pain interference items and shows relevance postoperatively. This is consistent with the finding in the mBPI study reported by Mendoza et al.8 The study also provides evidence supporting the reliability, validity, and responsiveness of the mBPI. These findings are consistent with and extend those reported in Mendoza et al,8 who studied only a CABG sample. The only substantive differences between the mBPI and the mBPI-e are the addition of the least pain severity item and the coughing and breathing items. As demonstrated in previous studies, the least pain item might not add substantially to the assessment of pain intensity, and many studies focus on either worst or average pain.8 We found that that the addition of the coughing and deep breathing items to the pain interference did not consistently improve the psychometric qualities of the mBPI-e. In this regard, both the mBPI and mBPI-e may be recommended for studies comparing treatments for postoperative pain in hospital settings. Clinically, the mBPI-e, with the added coughing and breathing items, might prove useful for understanding pain interference outcomes for some surgeries, such as GI and CABG. By using mBPI-e, physicians may be able to better assess patients’ postoperative pain where there is impact on the chest and abdomen. On the other hand, BPI might be recommended for studies in more general settings where normal daily activities are relevant because of the “normal work” and “enjoyment of life” items. The mBPI-e can be a useful pain interference measure for clinical trials and other clinical studies. In clinical practice, this instrument may be used to track outcomes in hospitalized patients after surgery. In addition, the mBPI-e may be helpful for comparative effectiveness studies examining different surgical interventions and different pain management regimens post surgery. These types of comparative effectiveness studies may assist clinicians and health policy-makers in determining those surgical and postsurgical interventions that improve patient outcomes and maximize efficient medical care. In summary, this study of the item characteristics and psychometric qualities of the mBPI-e provides evidence supporting the reliability, validity, and responsiveness of the pain severity and interference scales. The findings indicate good internal consistency and test-retest reliability, evidence supporting construct validity and responsiveness to changes in pain over the study duration. Effect sizes were large for pain severity and interference based on pain diary scores. The mBPI-e has been demonstrated to be a valid instrument that can be used to assess postoperative acute pain, perhaps specifically for surgical procedures with impact on the chest and abdomen. Although this study did not focus on determining the minimal important difference for the mBPI-e scales, the results suggest that a 0.62- to 0.89-point difference in pain severity and a 0.40- to 0.76point difference in pain interference scores might be clinically meaningful. The strength of this study was that it used the data of 2 large clinical trails with a combined sample size of well over 2000, with varieties of surgery. This assured the generalizability of the study results. The limitation of this study was the findings of the poor model fit to the 2-factor and the weakly good fit to the 3-factor construct of these 12 items based on the CFI. We believe the model fit will improve if items with high correlation between themselves will be deleted. However, content validity needs to be re-assessed if any item is to be deleted. In addition, further research is needed to confirm these preliminary estimates of minimal important differences in mBPI-e scores. 1
Current affiliation: Merck & Co., North Wales, Pennsylvania. Corresponding Author: Wen-Hung Chen, PhD, United BioSource Corporation, 7101 Wisconsin Avenue, Suite 600, Bethesda, MD 20814. E-mail address: [email protected] The clinical trial data used in this work were provided by Pfizer, who supported and conducted the 2 clinical trials. The BPI and mBPI are copyrighted by Charles Cleeland at M.D. Anderson Cancer Center, the University of Texas,
e27
e28
Modified Brief Pain Inventory-Exploratory Form
Houston. For those interested in using the mBPI-e in their research, permission should be secured from Dr. Cleeland, given that the mBPI-e was developed based on the mBPI instrument. Conflicts of interest: The work reported here was performed under contract for Pfizer by United BioSource Corporation. Authors WHC, RH, and DR work for the United BioSource Corporation, which performed this work under contract. Author KSC, faculty of the Department of Health Policy and Management, Johns Hopkins Bloomberg School of Public Health, is a contracted consultant to United BioSource Corporation. TJG, faculty of the Department of Anesthesiology, Duke University Medical Center, is an external reviewer with no contractual arrangement with either United BioSource Corporation or Pfizer. CC is an employee of Pfizer. ML was an employee of Pfizer at the time this manuscript was written.
R E F E R E N CE S 1. Breitbart W, McDonald MV, Rosenfeld B, et al. Pain in ambulatory AIDS patients. I: Pain characteristics and medical correlates. Pain. 1996;68(2-3): 315-321. 2. Cleeland C. Research in cancer pain. What we know and what we need to know. Cancer. 1991;67(3 Suppl):823-827. 3. Du Pen SL, Du Pen AR, Polissar N, et al. Implementing guidelines for cancer pain management: results of a randomized controlled clinical trial. J Clin Oncol. 1999;17(1):361-370. 4. Lydick E, Epstein RS, Himmelberger D, White CJ. Area under the curve: a metric for patient subjective responses in episodic diseases. Qual Life Res. 1995;4(1):41-45. 5. Tyler EJ, Jensen MP, Engel JM, Schwartz L. The reliability and validity of pain interference measures in persons with cerebral palsy. Arch Phys Med Rehabil. 2002;83(2):236-239. 6. Ward SE, Gordon D. Application of the American Pain Society quality assurance standards. Pain. 1994;56(3):299-306. 7. Zalon ML. Comparison of pain measures in surgical patients. J Nurs Meas. 1999;7(2):135-152.
8. Mendoza TR, Chen C, Brugger A, et al. The utility and validity of the modified brief pain inventory in a multiple-dose postoperative analgesic trial. Clin J Pain. 2004;20(5):357-362. 9. Nussmeier NA, Whelton AA, Brown MT, et al. Safety and efficacy of the cyclooxygenase-2 inhibitors parecoxib and valdecoxib after noncardiac surgery. Anesthesiology. 2006;104(3):518-526. 10. Ott E, Nussmeier NA, Duke PC, et al. Efficacy and safety of the cyclooxygenase 2 inhibitors parecoxib and valdecoxib in patients undergoing coronary artery bypass surgery. J Thorac Cardiovasc Surg. 2003;125(6):1481-1492. 11. Samejima F. Estimation of latent trait ability using a response pattern of graded scores. Psychometrika Monogr. 1969;(4):34. Pt 2: No. 17. 12. Thissen D. MULTILOG 7.0 User’s Guide. Lincolnwood, IL: Scientific Software International; 2002. 13. Deyo RA, Diehr P, Patrick DL. Reproducibility and responsiveness of health status measures. Statistics and strategies for evaluation. Control Clin Trials. 1991;12(4 Suppl):142S-158S. 14. Kazis LE, Anderson JJ, Meenan RF. Effect sizes for interpreting changes in health status. Med Care. 1989;27(3 Suppl):S178-S189.
Validation of the Modified Brief Pain Inventory-Exploratory Form in Surgery Patients Wen-Hung Chen, PhDa, Kitty S. Chan, PhDb, Tong J. Gan, MDc, Connie Chen, PharmDd, Mani Lakshminarayanan, PhDd,1, Dennis A. Revicki, PhDa a
CENTER FOR HEALTH OUTCOMES RESEARCH, UNITED BIOSOURCE CORPORATION, BETHESDA, MARYLAND; DEPARTMENT OF HEALTH POLICY AND MANAGEMENT, JOHNS HOPKINS BLOOMBERG SCHOOL OF PUBLIC HEALTH, BALTIMORE, MARYLAND; cDEPARTMENT OF ANESTHESIOLOGY, DUKE UNIVERSITY MEDICAL CENTER, DURHAM, NORTH CAROLINA; AND dPFIZER, NEW YORK, NEW YORK
b
-
ABSTRACT
O B J E C T I V E : An exploratory version of the Modified Brief Pain Inventory (mBPI-e) to measure acute post-operative pain, with new items on coughing, breathing, and concentration, was examined for their measurement properties. S T U D Y D E S I G N : This is a secondary study using data from two randomized clinical trials: general surgery trial (N=1050) and coronary artery bypass graft (CABG) surgery trial (N=1636). The measurements used in the two trials were: 1) mBPI-e; 2) clinician and patient global evaluations of medications; and 3) pain intensity diary. The mBPI-e and pain intensity were collected for 10 days. Clinician and patient global evaluations of medication were collected twice. The analyses conducted were: 1) exploratory factor analysis (EFA); 2) confirmatory factor analysis (CFA); 3) item response theory (IRT); 4) internal consistency; 5) test-retest reliability; 6) concurrent validity; 7) known-group validity; and 8) responsiveness. R E S U L T S : Pain severity, pain interference, and coughing and breathing factors were identified. Pain severity and pain interference subscale scores were constructed for mBPI-e. IRT analyses showed all items exhibited good item characteristics. Internal consistency was 0.85 for severity and 0.87 for interference. Test-retest reliability was 0.81 for severity and 0.71 for interference. Both severity and interference scores were correlated with diary-based pain intensity ratings (P <.0001). Mean severity and interference scores varied by physician and patient global ratings (P <.05). Severity and interference scores were responsive to changes in pain diary scores and physician global ratings (P <.001). There were no substantive differences in reliability or validity for sub-samples of surgery patients. C O N C L U S I O N S : The original BPI has been used in clinical studies, and the mBPI has demonstrated good reliability and validity in CABG patients. Based on this study, the mBPI-e has also demonstrated good reliability and validity for assessing postoperative acute pain in CABG and general surgery patients. See funding, conflict of interest, and authorship disclosures at the end of this article. Ó 2010 Elsevier Inc. All rights reserved Health Outcomes Research in Medicine (2010) 1, e17-e28
e18
Modified Brief Pain Inventory-Exploratory Form
K E Y W O R D S : Brief pain inventory; Measurement of pain; Modified brief pain inventory; Pain measure; Post operative pain
Postoperative
pain is common after highly invasive procedures such as coronary artery bypass graft (CABG) surgery, abdominal surgery, major orthopedic surgery, and gynecologic surgery. Clinical trials comparing different analgesic medications require psychometrically sound assessments of pain severity and interference with everyday activities. The Brief Pain Inventory (BPI) has been used in numerous studies and clinical trials of chronic and acute pain.1-7 In the postsurgical setting, ability to engage in normal daily activities is limited, and general pain assessments such as the pain interference scale from the BPI2 may not be entirely applicable. Mendoza et al8 developed a modified version of the BPI (mBPI) in CABG surgery patients, where items on general activity, normal work, and enjoyment of life were removed from the BPI and a new item on concentration was added. They found good evidence for internal consistency and test-retest reliability, confirmed the 2-domain factor structure and evidence supporting concurrent validity of the mBPI.
The objective of the current study was to evaluate the psychometric characteristics (ie, reliability, validity, responsiveness) of an exploratory modified version of the mBPI using data collected in two postoperative pain clinical trials. This study is the secondary data analysis of these 2 safety and efficacy trials. The exploratory version of the mBPI (mBPI-e) included 2 new items on interference related to coughing and deep breathing, which were intended to make the mBPI more relevant for assessing pain outcomes in postoperative analgesic clinical trials. The mBPI-e is based on the mBPI,8 and retains the original 4 pain severity items from the original BPI (worst, least, and average pain in the past 24 hours, and pain right now). For the new pain F I G U R E 1 : Items included in the BPI, mBPI, and mBPI-e. interference scale, mBPI-e retains the original general activity item and added 2 new items on coughing and deep breathing. These new items were added aiming to enhance the instrument’s ability in assessing acute pain outcomes in postoperative clinical studies. The assumption was that the interference associated with coughing and deep breathing can be very significant in many postoperative patients, depending on the type of surgery (eg, CABG, gastrointestinal [GI]). Figure 1 compares the items contained in BPI, mBPI, and mBPI-e.
METHODS Data for this study were taken from 2 clinical studies with different inclusion/exclusion criteria. Study 069 was a randomized, double-blind, parallel-group, multicenter study of the safety and efficacy of parecoxib followed by valdecoxib compared with placebo in general surgery patients for treatment of postoperative pain. It was conducted in 113 centers in 14 countries from September 2002 to February 2003.9 Study 071 was a randomized, double-blind, parallel-group, multicenter study of the safety and efficacy of parecoxib/valdecoxib and placebo/valdecoxib compared with placebo for treatment of postoperative pain in patients who had CABG. It was conducted at 175 centers in 27 countries from January 2003 to January 2004.10 Both Phase
Health Outcomes Research in Medicine - Vol. 1 / No. 1 / July 2010
III clinical trials were conducted with 10 days of treatment and 30 days of follow-up after study drug administration was completed.
Measures Modified BPI-e: The mBPI-e was administered each night at bedtime, with a 24-hour recall starting the day after surgery to examine pain severity (worst, least, average, and current pain) and pain interference (general activity, mood, walking ability, relations with others, sleeping, coughing, deep breathing, and concentration). Responses for these mBPI-e items were captured through the use of 11epoint numeric rating scales anchored at 0 (no pain or does not interfere) and ranged through 10 (pain as bad as you can imagine or completely interferes). The mBPI-e yielded 2 scale scores: pain severity and pain interference. The 2 scale scores were computed as the means of their corresponding item ratings. The higher scores indicate greater pain intensity or increased functional impairment. Imputation of missing item responses was based on the 50% rule, that is, if there were more than 50% missing item responses, the score for that scale was set to missing. When < 50% of item responses were missing, we used the average of the nonmissing item responses to impute missing item responses. The scale score then was calculated as the mean of the nonmissing item responses. There were very few missing item response data, and no more than 0.24% of the total sample required imputed pain severity or interference scale scores.
Global Evaluation of Treatment: Patient and physician global evaluations of study medication were collected through the use of a 4-point scale (1 ¼ poor, 2 ¼ fair, 3 ¼ good, 4 ¼ excellent). Physicians recorded their global evaluations when they examined the patients at the time of patient transition from parenteral to oral study medication (IV/PO), and at the end of the study (or early termination). The patients also were asked to record their own global evaluations at the time of IV/PO transition, and at the end of study (or early termination) when they returned to the clinical site for a final visit.
Daily Pain Diary: Patients recorded their current pain level in a diary through the use of a verbal rating scale (0 ¼ none, 1 ¼ mild, 2 ¼ moderate, 3 ¼ severe) each day from baseline (Day 1) through Day 9 prior to the AM dosing (usually at 8:00 AM) and at 2 (ie, 10:00 AM), 4 (12:00 PM), 8 (4:00 PM), and 12 (8:00 PM) hours (prior to PM dosing) and at 24 hours (8:00 AM the next day, prior to AM dosing). Average pain intensity (API) was calculated by using the trapezoid rule to obtain the sum across the hours, then average over the hours between the first and the last valid ratings, and had a range of 0 to 3.
Analysis Populations The psychometric analyses used all intention-to-treat patients who received at lease 1 dose of the study medication. The pooled general and CABG surgery data were structured by numbered study days. For consistency with the effect of the medication to the level of pain, this study designated baseline measurement as the day after the surgery.
Psychometric Analyses One focus of this evaluation study was the inclusion of the 2 new items (ie, coughing and deep breathing) in the pain interference scale. Therefore, we examined whether these 2 new items measure the same construct as the other original pain interference items, and whether they add additional validity to the instrument. We also assessed the item performance separately for the CABG and GI subgroups, and the nonGI subgroup, with focus on the coughing and deep breathing items. This was because these 2 items were more specific to the CABG and GI patients, and the item characteristics might be different between the 2 subgroups.
Item Analysis: We assessed mBPI-e item performance using data at baseline. We conducted exploratory factor analysis and confirmatory factor analysis to assess the dimensionality. We conducted item response
e19
e20
Modified Brief Pain Inventory-Exploratory Form
theory (IRT) analysis to assess whether the item response categories were monotonically ordered (ie, from lowest response option to highest response option) for each item and whether each item associated strongly with the underlying construct (ie, pain). For the factor analysis and IRT analysis, the sample was randomly divided into 2 subgroups. Exploratory factor analysis was conducted using 1 subgroup, and the confirmatory factor analysis was performed using the other subgroup. Promax rotation was applied to the factor structure, allowing the derived factors to be correlated. Eigenvalue > 1 was used as the criterion to extract factors. For the IRT analyses we used the Samejima graded response model11 and MULTILOG.12
Reliability: After the analysis of the item characteristics and item performance, we determined the final content of the mBPI-e. We then assessed the internal consistency reliability (Cronbach alpha) using baseline data. We assessed the test-retest reliability using data from Day 5 to Day 9 with a stable group of patients. Stable patients were defined as no change in their API scores within these assessment intervals. Intraclass correlation coefficients (ICCs) based on analysis of variance models were used to assess test-retest reliability.13
Construct Validity: We assessed the convergent validity by correlating the mBPI-e scale scores with the average rating of pain intensity for baseline, Day 3, Day 5, and Day 9. We assessed the known group validity of the instrument by comparing the mean mBPI-e scale scores across the classifications by the physician’s global evaluation of study medication and by the patient’s global evaluation of study medication.
Responsiveness: We assessed the responsiveness of the instrument by comparing the score change between responders and nonresponders from baseline to Day 3, and from baseline to Day 9 in the overall sample. Two definitions of responders and nonresponders were created, one based on the API, and the other by the physician global evaluation of study medication. Using the API, we defined the responder as patients with at least 1 point decrease in API scores from baseline to the target study day. This also corresponds to at least 1 level lower in pain intensity. A half-point F I G U R E 2 : Days of data used for psychometric decrease in API was defined as analyses. a borderline responder, and all other patients were defined as nonresponders. Analysis of covariance was conducted, testing mean score changes among the 3 groups, controlling for gender, age, and mBPI-e score at baseline. Based on the physician’s global evaluation, responders were defined as those patients who had good or excellent ratings, and all other patients were considered nonresponders. Analysis of covariance models, adjusting for gender, age, and mBPI-e score at baseline, were conducted to compare mean score differences for responders and nonresponders. Finally, the effect size by responders and nonresponders based on the API was estimated. Effect size was calculated as the difference in raw mean scores for responders and for nonresponders from baseline to endpoint (Day 3 or Day 9) divided by the standard deviation of the scale score at baseline.14 In all the analyses, nominal P-values were reported and no adjustments were made for any multiplicity. Statistical significance was claimed when P-value was <.05. Figure 2 shows the days of the data used for each of the analyses above.
Health Outcomes Research in Medicine - Vol. 1 / No. 1 / July 2010
e21
RESULTS Patient Demographics In the combined sample, mean age was 58.5 years (standard deviation [SD] ¼ 12.1) and the group was mostly male (68.4%) and white (93.2%) (Table 1). There were some differences between the general surgery and the CABG studies, with older patients (mean 61.9 vs 53.3 years) and more males in the CABG study.
Item Analyses Item Characteristics and Correlations: For the pain intensity and interference items, the entire range of item responses was observed. For pain intensity items, there was no evidence of floor or ceiling effects. For pain interference items, there was no evidence of a ceiling effect (range 1%-10%), but some evidence of a floor effect (range 11%-33%). At baseline, for pain interference, the highest reported interference was on general activity (mean 4.6, SD 3.3) and walking ability (mean 4.6, SD 3.8), and the least reported interference was on relations with others (mean 2.0, SD 2.7) and concentration (mean 2.2, SD 2.8). The coughing and breathing items were more likely to be endorsed by the GI and CABG surgery group (coughing: mean 3.9, SD 3.4; breathing: mean 4.0, SD 3.2) compared with the non-GI surgery group (coughing: mean 2.5, SD 3.3; breathing: mean 1.7, SD 2.6). The inter-item correlations ranged from 0.41 (worst and least pain) to 0.73 (worst and average pain) within pain severity, and ranged from 0.24 (sleep and coughing) to 0.67 (general activity and walking ability) within pain interference. The correlations between the pain severity and interference items ranged from 0.10 (least pain and coughing) to 0.52 (worst pain and general activity).
Factor Analysis: The exploratory factor analysis was conduced using one random split-half sample. Promax rotation was applied to the factor structure to allow the factors to be correlated. The results supported a 3factor solution with eigenvalues of 5.66, 1.57, and 1.10. Further inspection of the factor loading found that - T A B L E 1 : Demographics for Modified Intent-to-Treat Population* (n ¼ 2686) from General Surgery Study 069 (n ¼ 1050) and CABG Study 071 (n ¼ 1636)
Age (years), mean (SD) Gender, n (%) Female Male Race, n (%) White Black Asian Not listed Surgery type, n (%) Orthopedic Gastrointestinal Gynecologic Thoracic CABG Other surgery not listed
Combined
General Surgery Study
58.5 (12.13)
53.3 (14.35)
848 (31.6%) 1838 (68.4%)
617 (58.8%) 433 (41.2%)
231 (14.1%) 1405 (85.9%)
2504 (93.2%) 77 (2.9%) 58 (2.2%) 47 (1.7%)
977 (93.0%) 47 (4.5%) 3 (0.3%) 23 (2.2%)
1527 (93.3%) 30 (1.8%) 55 (3.4%) 24 (1.5%)
279 (10.4%) 391 (14.6%) 204 (7.6%) 22 (0.8%) 1636 (60.9%) 199 (7.4%)
279 (26.6%) 391 (37.2%) 204 (19.4%) 22 (2.1%) 0 (0.0%) 199 (19.0%)
0 (0.0%) 0 (0.0%) 0 (0.0%) 0 (0.0%) 1636 (100.0%) 0 (0.0%)
CABG ¼ coronary artery bypass graft; SD ¼ standard deviation. *Patients randomized to treatment who received at least 1 dose of study medication.
CABG Study 61.9 (8.96)
e22
Modified Brief Pain Inventory-Exploratory Form
factors could be identified as pain severity, pain interference, and a third factor consisting of the breathing and coughing items (Table 2). Given that the coughing and breathing items are more relevant for the GI and CABG patients, the factor analysis was repeated, including only the GI and CABG surgery patients. In this factor analysis, only 2 factors were extracted (eigenvalues: 6.24 and 1.42): pain severity and pain interference, and the coughing and breathing items were in the pain interference factor. The confirmatory factor analysis was conducted using the other split-half sample with 2-factor structure: 4 items for pain severity and 8 items for pain interference. Bentler’s comparative fit index (CFI) was 0.859, the standardized root mean residual (SRMR) was 0.067, and the root mean square error of approximation (RMSEA) was 0.129 (90% confidence interval, 0.122-0.137). These statistics did not meet the acceptable good-fit criteria (>0.90 for CFI, and <.05 for SRMR and RMSEA). The confirmatory factor analysis also was conducted using the 3-factor structure: 4 items for pain severity, 6 items for pain interference, and 2 coughing and breathing items. CFI was 0.910, the SRMR was 0.050, and the RMSEA was 0.105 (90% confidence interval, 0.097-0.113). A good fit to the observed data was accepted based on CFI and SRMR having met the criteria, even though RMSEA had not. We believe that all 3 fit statistics would reach the acceptable range if we 1) increase the number of factors; or 2) reduce the number of items. On the other hand, evidence of eigenvalues and factor loadings (Table 2) discouraged increase in the number of factors. And the content coverage (in the light of content validity) discouraged deleting the items. Therefore, we decided to accept the factor structure and acknowledge that the RMSEA did not meet the good-fit criterion.
Item Response Theory Analysis: We conducted an IRT analysis to assess the psychometric properties of the items. For pain severity, the 4 items covered the range of the pain severity domain. However, the average pain item has a substantially large slope parameter (7.58), which might be due to redundancy with the other 3 pain severity items (2.01, 2.02, 2.18). For pain interference, slope parameters ranged from 0.90 (for coughing) to 2.88 (for mood). Separate IRT analysis for the other split-half sample yielded similar findings.
Item Review Summary: Based on the item analyses, factor analyses, and the IRT analyses, the mBPI-e clearly contains 2 domains: pain severity and pain interference. Based on the IRT analyses of the pain severity items, fewer items are sufficient to assess pain severity or intensity. However, from the content and face validity perspective and for consistency with the original BPI, all 4 items were retained.
- T A B L E 2 : Rotated Factor Pattern* of mBPI-e from Exploratory Factor Analysis Combined Group with Split-Half Sample (n ¼ 1340)†
Relations with others (Interference) Mood (Interference) Concentration (Interference) Sleep (Interference) Walking ability (Interference) General activity (Interference) Average (Severity) Least pain (Severity) Pain right now (Severity) Worst pain (Severity) Coughing (Interference) Breathing (Interference)
Factor 1
Factor 2
Factor 3
0.854 0.833 0.738 0.734 0.732 0.715 0.038 0.110 0.081 0.142 0.084 0.120
0.062 0.062 0.088 0.058 0.045 0.154 0.910 0.878 0.790 0.679 0.013 0.010
0.009 0.038 0.198 0.040 0.053 0.034 0.009 0.015 0.004 0.043 0.950 0.848
mBPI-e ¼ Modified Brief Pain Inventory-exploratory form. *Standardized Regression Coefficients after PROMAX oblique rotation. † Of 1340 observations, 525 are omitted due to missing values; 815 actually used in analysis.
Health Outcomes Research in Medicine - Vol. 1 / No. 1 / July 2010
For the pain interference scale, the results show that the item on concentration does measure the same construct as with other pain inference items. The coughing and breathing items were more relevant for the GI and CABG patients than with other general surgery patients. These items do not consistently separate into a third factor, and it was uncertain whether these new items add to the measurement of pain interference in all patients. Therefore, we conducted the following psychometric analyses based on 3 sets of data: 1) the combined sample was analyzed with and without the coughing and breathing items; 2) GI and CABG patient samples were analyzed with and without the coughing and breathing items; and 3) the other general surgery patient subset was analyzed, with and without the coughing and breathing items. The intent was to compare the reliability and validity by different subgroups with and without these 2 items to evaluate their contribution to assessing pain interference.
Reliability Internal Consistency Reliability, and Test-retest Reliability: The internal consistency reliability, at baseline, was 0.85 for the pain severity scale and 0.87 for the pain interference scale (with or without the coughing and breathing items). For the GI and CABG patients, internal consistency was 0.89 with the coughing and breathing items and 0.88 without these items. For the other general surgery patients, internal consistency reliability was 0.83 and 0.85 with and without the coughing and breathing items, respectively. Test-retest reliability for pain severity was 0.81. For pain interference, the ICCs were 0.75 and 0.73 for the overall sample with and without the coughing and breathing items, respectively. For the GI and CABG patients, ICCs for pain interference were 0.74 with the coughing and breathing items and 0.71 without the 2 items. For the general surgery group, ICCs were 0.81 and 0.78 with and without the 2 items, respectively.
Validity Construct Validity: Construct validity for mBPI-e was assessed by the correlations between the pain severity and interference scale scores and the API for Days 1, 3, 5, and 9. As expected, the correlation between the pain severity and API is higher than the correlation between the pain inference and API. The correlations between the pain severity scale score and the API ranged from 0.57 to 0.67 over these assessments (all P <.001). The correlations between the pain interference scale score and the API ranged from 0.39 to 0.49 (all P <.001). The correlations between pain interference and API with or without the coughing and breathing items were not substantively different (data not shown). Known-group Validity: Known-group validity assesses the ability of the mBPI-e scale scores to distinguish between different groups of patients. For this study, we based the examination of known-group validity on the physician’s global evaluation of study medication (Table 3) and the patient’s global evaluation of study medication (Table 4). Based on the physician’s global evaluation, at IV/PO switch, both mBPI-e scale scores were statistically significantly different between the 4 groups (all P <.001), except between poor and fair categories (P >.05). At end of study, the mBPI-e scale scores were significantly different (all P <.05), except between good and excellent groups (P >.05). Based on the patient’s global evaluation, the mBPI-e scale scores were significantly different between every pair of groups at IV/PO switch (all P <.001). At end of study, the mBPI-e scale scores were significantly different (all P <.05), except those between fair and good groups (P >.05), and between good and excellent groups for pain interference (P >.05). Overall, both mBPI-e scale scores show very good known-group validity in terms of mean score differences based on the physician global evaluations of study medication. These findings were not substantively different with or without the coughing or breathing items, or by surgical subgroup (data not shown).
Responsiveness: API-defined responders reported significantly greater reductions in pain severity, with mean decreases of 1.93 points, compared to mean decreases of 0.80 point for nonresponders (P <.0001) and
e23
e24
Modified Brief Pain Inventory-Exploratory Form
- T A B L E 3 : Mean mBPI-e Scores by Physician Global Evaluation of Study Medication at IV/PO Switching and End of Study Modified Intent to Treat Population (n ¼ 2686) Physician Global Evaluation of Study Medication mBPI-e Scales At IV/PO switch Pain severity Pain interference At end/early termination of study Pain severity Pain interference
Poor n, Mean (SD)
Fair n, Mean (SD)
Good n, Mean (SD)
Excellent n, Mean (SD)
P-Value*
122, 3.7 (1.92) 122, 3.2 (2.06)
281, 3.3 (1.77) 284, 2.7 (2.00)
600, 2.7 (1.66) 603, 2.1 (1.82)
328, 2.0 (1.41) 327, 1.6 (1.60)
2§,3§,4§,5§,6§ 2§,3§,4‡,5§,6‡
72, 3.6 (2.56) 72, 2.8 (2.79)
99, 2.7 (1.90) 99, 1.9 (2.05)
204, 1.9 (1.64) 205, 1.3 (1.51)
127, 1.4 (1.41) 127, 0.9 (1.11)
1†,2§,3§,4†,5§ 1†,2§,3§,4†,5‡
IV ¼ intravenous (parenteral); mBPI-e ¼ Modified Brief Pain Inventory-exploratory form; PO ¼ oral; SD ¼ standard deviation. *One-way analysis of variance using Scheffe’s post hoc comparisons for differences between groups; 1 ¼ poor vs fair; 2 ¼ poor vs good; 3 ¼ poor vs excellent; 4 ¼ fair vs good; 5 ¼ fair vs excellent; 6 ¼ good vs excellent. † P <.05. ‡ P <.001. § P <.0001.
compared with borderline responders (1.93 vs 1.38, P <.05) at Day 3 (Table 5). Comparable significant differences were observed in mean changes in pain severity scores between responders and nonresponders (2.90 vs 1.39, P <.001) and responders and borderline responders (2.90 vs 2.28, P <.001) at Day 9. Responders reported significantly more reductions in pain interference compared with nonresponders (1.84 vs 1.02, P <.05) on Day 3. By Day 9, API-based responders reported significant differences on pain inference compared with borderline responders (3.03 vs 2.63, P <.05) and with nonresponders (3.03 vs 1.87, P <.0001). At IV/PO switch, physician global rating-defined responders reported mean decreases of 1.23 points in pain severity scores compared with mean decreases of 0.69 points in nonresponders (P <.0001) (Table 6). At end of study, responders reported significantly greater decreases in pain severity scores compared with nonresponders (2.22 vs 1.33, respectively, P <.0001). Also, responders reported significantly greater
- T A B L E 4 : Mean of mBPI-e Scores by Patient Global Evaluation of Study Medication at IV/PO Switching and End of Study Modified Intent to Treat Population (n ¼ 2686) Physician Global Evaluation of Study Medication mBPI-e Scales At IV/PO switching Pain severity Pain interference At end/early termination of study Pain severity Pain interference
Poor n, Mean (SD)
Fair n, Mean (SD)
Good n, Mean (SD)
Excellent n, Mean (SD)
P-Value*
93, 4.2 (1.92) 93, 3.7 (2.19)
258, 3.2 (1.61) 260, 2.6 (1.99)
651, 2.7 (1.69) 655, 2.1 (1.76)
328, 2.1 (1.58) 326, 1.6 (1.65)
1‡,2§,3§,4‡,5§,6§ 1§,2§,3§,4‡,5§,6‡
72, 4.2 (2.49) 72, 3.4 (2.97)
81, 2.5 (1.65) 81, 1.7 (1.44)
227, 1.9 (1.63) 227, 1.2 (1.43)
120, 1.3 (1.33) 121, 0.7 (1.02)
1§,2§,3§,5§,6† 1§,2§,3§,5‡
IV ¼ intravenous (parenteral); mBPI-e ¼ Modified Brief Pain Inventory-exploratory form; PO ¼ oral; SD ¼ standard deviation. *One-way analysis of variance using Scheffe’s post hoc comparisons for differences between groups; 1 ¼ poor vs fair; 2 ¼ poor vs good; 3 ¼ poor vs excellent; 4 ¼ fair vs good; 5 ¼ fair vs excellent; 6 ¼ good vs excellent. † P <.05. ‡ P <.001. § P <.0001.
Health Outcomes Research in Medicine - Vol. 1 / No. 1 / July 2010
e25
- T A B L E 5 : The Responsiveness of mBPI-e by Average Pain Intensity (Responder vs Nonresponder) mITT Population (n ¼ 2686) Classification by change in Average Pain Intensity* ‡
mBPI-e Scales From baseline to day 3 Pain severity Pain interference From baseline to day 9 Pain severity Pain interference
Responder n, LS Mean (SE)
Borderline Responder§ n, LS Mean (SE)
Nonresponderk n, LS Mean (SE)
Overall Model† P-Value
Pair Comparison{
137, 1.93 (0.125) 137, 1.84 (0.143)
280, 1.38 (0.087) 280, 1.45 (0.100)
832, 0.80 (0.052) 831, 1.02 (0.060)
<.0001 <.0001
1††,2††, 3# 1#, 2#
152, 2.90 (0.118) 152, 3.03 (0.107)
148, 2.28 (0.120) 148, 2.63 (0.110)
332, 1.39 (0.083) 334, 1.87 (0.076)
<.0001 <.0001
1††,2††, 3** 1††,2††, 3#
LS ¼ least square; mBPI-e ¼ Modified Brief Pain Inventory-exploratory form; mITT ¼ Modified Intent to Treat; SE ¼ standard error. *Average Pain Intensity (API) is computed using trapezoid rule on observed ratings such that API ¼ 0.5 {2 (PI0 þ PI2) þ 2 (PI2 þ PI4) þ 4 (PI4 þ PI8) þ 4 (PI8 þ PI12) þ 12 (PI12 þ PI24)}. † Analysis of covariance controlling for age, gender, and baseline mBPI-e scale score. ‡ API decreases more than 1 point. § API decreases more than 0.5 point but less than 1 point. k API decreases less than 0.5 point. { 1 ¼ nonresponder vs borderline responder; 2 ¼ nonresponder vs responder; 3 ¼ borderline responder vs responder. # P <.05. **P <.001. †† P <.0001.
reductions in pain interference compared with nonresponders (P <.05) at IV/PO switch. By end of study, responders reported mean decreases of 2.43 points compared with mean decreases of 1.75 points in nonresponders (P <.0001). There was no difference in the responsiveness of the interference scale with or without the coughing and breathing items.
- T A B L E 6 : The Responsiveness of mBPI-e by Physician Global Evaluation of Study Medication (Responder vs Nonresponder), mITT Population (n ¼ 2686) Physician Global Evaluation of Study Medication
mBPI-e Scales At IV/PO switching Pain severity Pain interference At end/early termination of study Pain severity Pain interference
Responder† n, LS Mean (SE)
Nonresponder‡ n, LS Mean (SE)
Overall Model P-Value*
Pair Comparison§
835, 1.23 (0.050) 836, 1.36 (0.056)
379, 0.69 (0.076) 380, 0.98 (0.083)
<.0001 <.0001
1{ 1k
309, 2.22 (0.098) 311, 2.43 (0.093)
167, 1.33 (0.135) 165, 1.75 (0.129)
<.0001 <.0001
1{ 1{
IV ¼ intravenous (parenteral); LS ¼ least square; mBPI-e ¼ Modified Brief Pain Inventory-exploratory form; mITT ¼ Modified Intent to Treat; PO ¼ oral; SE ¼ standard error. *Analysis of covariance controlling for age, gender, and baseline mBPI-e scale score. † Global evaluation Good or Excellent. ‡ Global evaluation Poor or Fair. § Responder vs Nonresponder. k P <.001. { P <.0001.
e26
Modified Brief Pain Inventory-Exploratory Form
Effect Size: For pain severity response based on the API at Day 3, the effect size for responders was 1.25 and for nonresponders was 0.37. The effect sizes for pain interference at Day 3 were lower, but different among the 3 responder groups. Effect sizes for pain interference were 0.93, 0.68, and 0.39 for the responder, borderline responder, and nonresponder groups, respectively. At Day 9, the responder group’s effect size for pain severity was 1.81, for borderline responders it was 1.33, and for nonresponders it was 0.64. The effect sizes for pain interference at Day 9 were 1.52, 1.22, and 0.70 for the responder, borderline responder, and nonresponder groups, respectively.
DISCUSSION The data used in this study for the psychometric analysis of mBPI-e combined the clinical trial data that were reported in both Nussmeier et al9 and Ott et al,10 who studied the safety and analgesic efficacy of parecoxib and valdecoxib in comparison with placebo. The mBPI-e data were not reported in Nussmeier et al (2006; study 069),9 and only group comparison results for individual items were reported in Ott et al (2003; study 071).10 This study evaluated the item characteristics and psychometric qualities of the mBPI-e based on a secondary analysis of clinical trial data. The goal was to evaluate the reliability, validity, and responsiveness of the mBPI-e for assessing acute pain in a postoperative setting. Our study found that the mBPI-e is a reliable, valid, and responsive measure of postoperative pain, which could be especially useful for evaluating the effectiveness of the pain treatment for patients who have undergone surgery. Based on the item analyses, no problems were identified for any items that warranted deletion from the revised pain measure. The exploratory factor analyses supported a 3-factor (domain) structure when general, GI, and CABG surgery types of data were included in the analyses, with the new breathing and coughing items forming the third factor. This was likely attributable to the fact that the breathing and coughing items were not as relevant for the general surgery patients. When including only the GI and CABG surgery data, the exploratory and confirmatory factor analyses supported the 2-factor solution with pain severity and pain interference. In the GI and CABG patients, the breathing and coughing items contribute to the pain interference factor. The pain severity and pain interference scales both showed acceptable internal consistency and testretest reliability. Therefore, reliability for the pain severity and interference scales also was supported in this study. The pain severity and interference scales showed evidence supporting construct validity and discriminant validity in this sample of CABG and general surgery patients. As expected, pain severity and API diary scores were moderately and significantly correlated. There were significant correlations between the diarybased pain measures and pain interference, indicating that greater pain severity was associated with more interference in activities, walking, mood, and other aspects of function. Average pain severity and pain interference scores also varied significantly by patient-rated and clinician-rated global evaluations of treatment. For example, the mean pain severity scores for the excellent or good groups based on physician ratings were significantly better than those rated as poor at the end of the study (mean score 1.4 and 1.9 vs 3.6). Similar differences between physician-defined groups were observed for the pain interference scores as well, with the mean scores in the excellent group 0.9 points lower (better) than the good group, 1.5 points lower than the fair group, and 1.9 points lower than the poor group. The mean differences based on patient global ratings were comparable and supported the differences observed based on clinician ratings. Responsiveness was demonstrated for both the pain severity and pain interference scales related to changes in pain severity based on daily diaries and by physician global evaluations of treatment effects. By the end of the study, pain diary responders decreased their pain severity scores by 2.9 points, compared with decreases of 1.4 points in nonresponders. For pain interference, responders reported 3.0 point decreases (improvements), whereas nonresponders reported 1.9 point decreases. For clinical trials, evidence of responsiveness is important, and based on this study, the mBPI-e pain severity and interference scales are highly responsive to changes in daily pain ratings and physician global evaluations.
Health Outcomes Research in Medicine - Vol. 1 / No. 1 / July 2010
The mBPI-e contains 2 new items that extend the coverage of concepts for pain studies in the postoperative hospital setting. The new items on coughing and breathing were more relevant for GI and CABG patients than for other general surgery patients. However, extended and intensive analyses on reliability and validity show that the 2 items can be included or excluded in the instrument and do not effect the measurement properties of the pain interference scale, including responsiveness. Therefore, based on these findings, the recommendation was to include these 2 items in the interference scale when assessing GI and CABG patients, because they add content validity and cover additional areas relevant for GI and CABG surgery patients. The concentration item fits in with the other original pain interference items and shows relevance postoperatively. This is consistent with the finding in the mBPI study reported by Mendoza et al.8 The study also provides evidence supporting the reliability, validity, and responsiveness of the mBPI. These findings are consistent with and extend those reported in Mendoza et al,8 who studied only a CABG sample. The only substantive differences between the mBPI and the mBPI-e are the addition of the least pain severity item and the coughing and breathing items. As demonstrated in previous studies, the least pain item might not add substantially to the assessment of pain intensity, and many studies focus on either worst or average pain.8 We found that that the addition of the coughing and deep breathing items to the pain interference did not consistently improve the psychometric qualities of the mBPI-e. In this regard, both the mBPI and mBPI-e may be recommended for studies comparing treatments for postoperative pain in hospital settings. Clinically, the mBPI-e, with the added coughing and breathing items, might prove useful for understanding pain interference outcomes for some surgeries, such as GI and CABG. By using mBPI-e, physicians may be able to better assess patients’ postoperative pain where there is impact on the chest and abdomen. On the other hand, BPI might be recommended for studies in more general settings where normal daily activities are relevant because of the “normal work” and “enjoyment of life” items. The mBPI-e can be a useful pain interference measure for clinical trials and other clinical studies. In clinical practice, this instrument may be used to track outcomes in hospitalized patients after surgery. In addition, the mBPI-e may be helpful for comparative effectiveness studies examining different surgical interventions and different pain management regimens post surgery. These types of comparative effectiveness studies may assist clinicians and health policy-makers in determining those surgical and postsurgical interventions that improve patient outcomes and maximize efficient medical care. In summary, this study of the item characteristics and psychometric qualities of the mBPI-e provides evidence supporting the reliability, validity, and responsiveness of the pain severity and interference scales. The findings indicate good internal consistency and test-retest reliability, evidence supporting construct validity and responsiveness to changes in pain over the study duration. Effect sizes were large for pain severity and interference based on pain diary scores. The mBPI-e has been demonstrated to be a valid instrument that can be used to assess postoperative acute pain, perhaps specifically for surgical procedures with impact on the chest and abdomen. Although this study did not focus on determining the minimal important difference for the mBPI-e scales, the results suggest that a 0.62- to 0.89-point difference in pain severity and a 0.40- to 0.76point difference in pain interference scores might be clinically meaningful. The strength of this study was that it used the data of 2 large clinical trails with a combined sample size of well over 2000, with varieties of surgery. This assured the generalizability of the study results. The limitation of this study was the findings of the poor model fit to the 2-factor and the weakly good fit to the 3-factor construct of these 12 items based on the CFI. We believe the model fit will improve if items with high correlation between themselves will be deleted. However, content validity needs to be re-assessed if any item is to be deleted. In addition, further research is needed to confirm these preliminary estimates of minimal important differences in mBPI-e scores. 1
Current affiliation: Merck & Co., North Wales, Pennsylvania. Corresponding Author: Wen-Hung Chen, PhD, United BioSource Corporation, 7101 Wisconsin Avenue, Suite 600, Bethesda, MD 20814. E-mail address: [email protected] The clinical trial data used in this work were provided by Pfizer, who supported and conducted the 2 clinical trials. The BPI and mBPI are copyrighted by Charles Cleeland at M.D. Anderson Cancer Center, the University of Texas,
e27
e28
Modified Brief Pain Inventory-Exploratory Form
Houston. For those interested in using the mBPI-e in their research, permission should be secured from Dr. Cleeland, given that the mBPI-e was developed based on the mBPI instrument. Conflicts of interest: The work reported here was performed under contract for Pfizer by United BioSource Corporation. Authors WHC, RH, and DR work for the United BioSource Corporation, which performed this work under contract. Author KSC, faculty of the Department of Health Policy and Management, Johns Hopkins Bloomberg School of Public Health, is a contracted consultant to United BioSource Corporation. TJG, faculty of the Department of Anesthesiology, Duke University Medical Center, is an external reviewer with no contractual arrangement with either United BioSource Corporation or Pfizer. CC is an employee of Pfizer. ML was an employee of Pfizer at the time this manuscript was written.
R E F E R E N CE S 1. Breitbart W, McDonald MV, Rosenfeld B, et al. Pain in ambulatory AIDS patients. I: Pain characteristics and medical correlates. Pain. 1996;68(2-3): 315-321. 2. Cleeland C. Research in cancer pain. What we know and what we need to know. Cancer. 1991;67(3 Suppl):823-827. 3. Du Pen SL, Du Pen AR, Polissar N, et al. Implementing guidelines for cancer pain management: results of a randomized controlled clinical trial. J Clin Oncol. 1999;17(1):361-370. 4. Lydick E, Epstein RS, Himmelberger D, White CJ. Area under the curve: a metric for patient subjective responses in episodic diseases. Qual Life Res. 1995;4(1):41-45. 5. Tyler EJ, Jensen MP, Engel JM, Schwartz L. The reliability and validity of pain interference measures in persons with cerebral palsy. Arch Phys Med Rehabil. 2002;83(2):236-239. 6. Ward SE, Gordon D. Application of the American Pain Society quality assurance standards. Pain. 1994;56(3):299-306. 7. Zalon ML. Comparison of pain measures in surgical patients. J Nurs Meas. 1999;7(2):135-152.
8. Mendoza TR, Chen C, Brugger A, et al. The utility and validity of the modified brief pain inventory in a multiple-dose postoperative analgesic trial. Clin J Pain. 2004;20(5):357-362. 9. Nussmeier NA, Whelton AA, Brown MT, et al. Safety and efficacy of the cyclooxygenase-2 inhibitors parecoxib and valdecoxib after noncardiac surgery. Anesthesiology. 2006;104(3):518-526. 10. Ott E, Nussmeier NA, Duke PC, et al. Efficacy and safety of the cyclooxygenase 2 inhibitors parecoxib and valdecoxib in patients undergoing coronary artery bypass surgery. J Thorac Cardiovasc Surg. 2003;125(6):1481-1492. 11. Samejima F. Estimation of latent trait ability using a response pattern of graded scores. Psychometrika Monogr. 1969;(4):34. Pt 2: No. 17. 12. Thissen D. MULTILOG 7.0 User’s Guide. Lincolnwood, IL: Scientific Software International; 2002. 13. Deyo RA, Diehr P, Patrick DL. Reproducibility and responsiveness of health status measures. Statistics and strategies for evaluation. Control Clin Trials. 1991;12(4 Suppl):142S-158S. 14. Kazis LE, Anderson JJ, Meenan RF. Effect sizes for interpreting changes in health status. Med Care. 1989;27(3 Suppl):S178-S189.