Advertisement

Patterns of Pain and Interference in Patients with Painful Bone Metastases: A Brief Pain Inventory Validation Study

  • Jackson S.Y. Wu
    Correspondence
    Address correspondence to: Jackson S.Y. Wu, MD, Department of Oncology, Tom Baker Cancer Centre, University of Calgary, 1331–29th St NW, Calgary, Alberta T2N 4N2, Canada.
    Affiliations
    Department of Oncology, Tom Baker Cancer Centre, University of Calgary, Calgary, Alberta, Canada
    Search for articles by this author
  • Dorcas Beaton
    Affiliations
    Institute for Work & Health, Toronto, Ontario, Canada

    St. Michael's Hospital, University of Toronto, Toronto, Ontario, Canada
    Search for articles by this author
  • Peter M. Smith
    Affiliations
    St. Michael's Hospital, University of Toronto, Toronto, Ontario, Canada

    Dalla Lana School of Public Health, University of Toronto, Toronto, Ontario, Canada
    Search for articles by this author
  • Neil A. Hagen
    Affiliations
    Department of Oncology, Tom Baker Cancer Centre, University of Calgary, Calgary, Alberta, Canada

    Department of Clinical Neurosciences, Tom Baker Cancer Centre, University of Calgary, Calgary, Alberta, Canada

    Department of Medicine, Tom Baker Cancer Centre, University of Calgary, Calgary, Alberta, Canada

    Cancer Pain Clinic, Tom Baker Cancer Centre, University of Calgary, Calgary, Alberta, Canada
    Search for articles by this author

      Abstract

      Bone metastases are prevalent, painful, and carry a poorer prognosis for pain control compared with other cancer pain syndromes. Standard tools to measure pain have not been validated in this patient population, and particular subgroups with more challenging symptoms have yet to be identified and studied. The objectives of this study were 1) to validate the psychometric properties of the Brief Pain Inventory (BPI) and its Pain and Interference subscales in patients with clinically significant metastatic bone pain requiring palliative radiotherapy and 2) to examine differences in BPI subscales among predefined subgroups of bone metastases patients. A total of 258 patients evaluated and treated through a rapid access radiation therapy clinic between July 2002, and November 2006, were included in the analysis. High internal consistency of the BPI subscales of Pain, Activity interference, and Affect interference was demonstrated by Cronbach's alpha between 0.81 and 0.89. Removing sleep interference improved model fit in confirmatory factor analysis. The BPI revealed an alarming pattern in patients with lower body metastases, who reported substantial interference of activity even though pain levels were mild or moderate. Such patients may require prompt clinical attention to better meet their needs. Finally, the allocation of interference from sleep within the BPI framework, in our population of pain patients, requires further study.

      Key Words

      Introduction

      Pain is a highly prevalent symptom and a serious public health issue in patients with advanced, metastatic, or terminal cancer.
      • van den Beuken-van Everdingen M.H.
      • de Rijke J.M.
      • Kessels A.G.
      • et al.
      Prevalence of pain in patients with cancer: a systematic review of the past 40 years.
      Despite the prevalence of pain, a recent systematic review concluded that interventions to improve pain among hospitalized cancer patients fall far short of fully meeting patients' needs
      • Goldberg G.R.
      • Morrison R.S.
      Pain management in hospitalized cancer patients: a systematic review.
      and other reports have documented a similar undertreatment of pain among cancer outpatients.
      • Cleeland C.S.
      • Gonin R.
      • Hatfield A.K.
      • et al.
      Pain and its treatment in outpatients with metastatic cancer.
      • Cleeland C.S.
      The measurement of pain from metastatic bone disease: capturing the patient's experience.
      Bone is the most frequent site of spread in advanced malignancies such as cancers of the breast, prostate, and lung. Although fracture, hypercalcemia, and nerve entrapment are serious complications of bone metastases, pain is the most common clinical accompaniment and it is often difficult to control. Pain has been correlated with underlying bone destruction, but surprisingly, the various potential pathophysiological mechanisms of metastatic bone pain and their interconnections have yet to be fully delineated. The net result of unrelieved bone pain, however, is grim, frequently including diminished quality of life and increased reliance on pharmacologic or other therapeutic interventions.
      • Mercadante S.
      • Fulfaro F.
      Management of painful bone metastases.
      Accurate pain assessment techniques are an important tool to guide clinical decision making and evaluate treatment effectiveness in this heterogeneous population of patients with various stages of disease, mechanisms of pain, responses to analgesics, and comorbidities such as delirium, asthenia, and other contributors to high burden of illness. The Brief Pain Inventory (BPI) is a measurement tool that has demonstrated construct validity in two out of the four core outcome domains outlined by the Initiative on Methods, Measurement, and Pain Assessment in Clinical Trials (IMMPACT) consensus recommendations:
      • Dworkin R.H.
      • Turk D.C.
      • Wyrwich K.W.
      • et al.
      Interpreting the clinical importance of treatment outcomes in chronic pain clinical trials: IMMPACT recommendations.
      pain intensity and interference of function by pain.
      • Caraceni A.
      • Mendoza T.R.
      • Mencaglia E.
      • et al.
      A validation study of an Italian version of the Brief Pain Inventory (Breve Questionario per la Valutazione del Dolore).
      • Radbruch L.
      • Loick G.
      • Kiencke P.
      • et al.
      Validation of the German version of the Brief Pain Inventory.
      • Wang X.S.
      • Mendoza T.R.
      • Gao S.Z.
      • Cleeland C.S.
      The Chinese version of the Brief Pain Inventory (BPI-C): its development and use in a study of cancer pain.
      • Uki J.
      • Mendoza T.
      • Cleeland C.S.
      • Nakamura Y.
      • Takeda F.
      A brief cancer pain assessment tool in Japanese: the utility of the Japanese Brief Pain Inventory—BPI-J.
      • Saxena A.
      • Mendoza T.
      • Cleeland C.S.
      The assessment of cancer pain in north India: the validation of the Hindi Brief Pain Inventory—BPI-H.
      • Klepstad P.
      • Loge J.H.
      • Borchgrevink P.C.
      • et al.
      The Norwegian Brief Pain Inventory questionnaire: translation and validation in cancer pain patients.
      • Kalyadina S.A.
      • Ionova T.I.
      • Ivanova M.O.
      • et al.
      Russian Brief Pain Inventory: validation and application in cancer pain.
      Although most validation studies of the BPI confirm the legitimacy of these two factors, namely, pain intensity and interference of function, Cleeland et al.
      • Cleeland C.S.
      • Nakamura Y.
      • Mendoza T.R.
      • et al.
      Dimensions of the impact of cancer pain in a four country sample: new information from multidimensional scaling.
      have suggested that the interference factor can be further separated into subscales of Activity and Affect. Two subsequent studies have confirmed this three-factor structure.
      • Saxena A.
      • Mendoza T.
      • Cleeland C.S.
      The assessment of cancer pain in north India: the validation of the Hindi Brief Pain Inventory—BPI-H.
      • Klepstad P.
      • Loge J.H.
      • Borchgrevink P.C.
      • et al.
      The Norwegian Brief Pain Inventory questionnaire: translation and validation in cancer pain patients.
      Even though cancer-related bone pain and sometimes its treatment can seriously impair patients' functional ability, most studies have focused only on the construct validity of the pain-interference domains of the BPI, broadly grouping together cancer patients without investigating its ability to discriminate or characterize subgroups of cancer pain with respect to its anatomic origin. There is emerging evidence that some cancer pains predictably carry a poorer prognosis, such as breakthrough cancer pain. Breakthrough cancer pain is prevalent in patients with bone metastases.
      • Hagen N.A.
      • Biondo P.
      • Stiles C.
      Assessment and management of breakthrough pain in cancer patients: current approaches and emerging research.
      • Fainsinger R.L.
      • Nekolaichuk C.L.
      A “TNM” classification system for cancer pain: the Edmonton Classification System for Cancer Pain (ECS-CP).
      Based on our clinical experience within a rapid access radiotherapy clinic for bone pain,
      • Wu J.S.
      • Monk G.
      • Clark T.
      • et al.
      Palliative radiotherapy improves pain and reduces functional interference in patients with painful bone metastases: a quality assurance study.
      we have been suspicious that bone pain arising from the lower part of the body, including pelvis and lower extremity (lower skeletal pain), results in greater impairment of activity functions compared with bone pain arising from the upper body (upper skeletal pain), thereby requiring greater attention and proactive management. As such, an interaction may exist between the location of skeletal pain and functional interference, even when pain severity appears to be tolerable.
      Therefore, the present study has two main objectives: 1) to validate the psychometric properties of the BPI and its pain and interference subscales in patients with clinically significant metastatic bone pain requiring palliative radiotherapy and 2) to confirm our a priori hypothesis that lower skeletal metastases are associated with greater functional interference compared with upper body bone pain.

      Methods

      Population and Setting

      The Tom Baker Cancer Centre provides regional radiotherapy services to a catchment area of 1.8 million residents within southern Alberta. Approximately one hundred patients with painful bone metastases per year are referred to a specialized rapid access radiotherapy clinic for management. Most of the patients attending this clinic are managed as outpatients. The study population comprised all patients referred to this clinic treated between July 2002, and November 2006. This tool validation study was part of a quality assurance audit approved by the Conjoint Health Research Ethics Board, University of Calgary (ID #18955).

      Brief Pain Inventory

      The BPI consists of a body diagram, documentation of analgesia consumption, and 11 items on an ordinal scale from 0 to 10 in single-digit increments.
      • Cleeland C.S.
      The measurement of pain from metastatic bone disease: capturing the patient's experience.
      Two factors include pain perception and interference by pain. Pain perception is measured by four items: worst pain, average pain, least pain, and pain right now, which are anchored between “no pain” and “pain as bad as you can imagine.” Interference by pain is assessed by seven items anchored between “does not interfere” and “completely interferes”: general activity, mood, walking ability, normal work including both work outside the home and housework, relations with other people, sleeping, and enjoyment of life. In the studies that reported a three-factor solution, interference was divided into Activity interference (general activity, walking ability, and normal work) and Affect interference (mood, relations, and enjoyment). Sleep interference was included as an Activity item in one study
      • Cleeland C.S.
      • Nakamura Y.
      • Mendoza T.R.
      • et al.
      Dimensions of the impact of cancer pain in a four country sample: new information from multidimensional scaling.
      but as an Affect item in the other.
      • Klepstad P.
      • Loge J.H.
      • Borchgrevink P.C.
      • et al.
      The Norwegian Brief Pain Inventory questionnaire: translation and validation in cancer pain patients.
      An additional item of percentage pain relief is in the BPI questionnaire, but this item was not incorporated into analysis in prior validation studies.
      Two minor modifications to the BPI were made for our patient population: 1) the item “pain at its least” was deleted from the questionnaire as this item has been reported to carry poor discrimination
      • Daut R.L.
      • Cleeland C.S.
      • Flanery R.C.
      Development of the Wisconsin Brief Pain Questionnaire to assess pain in cancer and other diseases.
      and 2) patients were asked to rate their pain and interference items during “the past three days” rather than “last 24 hours.” This latter modification was made because an international consensus of endpoint measures for radiotherapy trials in bone pain recommended patients be asked to recall their pain experience over three days when scoring bone pain.
      • Chow E.
      • Wu J.S.
      • Hoskin P.
      • et al.
      International consensus on palliative radiotherapy endpoints for future clinical trials in bone metastases.
      The consensus statement reasoned that incidental or breakthrough pain is intermittent in nature and a longer recall period seems appropriate. We posit that these modifications would have no deleterious impact on the overall psychometric and clinical performance of the BPI instrument in this population of patients.

      Inclusion and Data Collection

      Patients with radiographically confirmed bone metastases who completed the English version of the BPI and subsequently underwent palliative radiotherapy were included in the study. Occasionally, patients required assistance by a dedicated clinic nurse or a family member to read the questionnaire, but most patients completed the questionnaires on their own within 10–15 minutes, before interview by physician. Patients who were too frail to fill out a questionnaire, who indicated zero pain on all three pain severity items, or those who completed the form partially (i.e., rated pain items but none of the interference) were excluded. Demographics and radiotherapy data, including anatomic site(s) of treatment and number of prescribed fractions, were collected at the time of treatment. For patients who had multiple evaluations and courses of bone radiotherapy over time, only the first available BPI evaluation was included in this analysis.
      For the purpose of examining discriminatory properties of the BPI, patient groups were predefined by location of clinically significant bone pain as treated by a single volume of radiotherapy. Patients who required radiotherapy to the lumbar spine, sacrum, or any of pelvic girdle (iliac wing, acetabulum, pubic bone, and ischial tuberosity), femur (head, neck, or shaft), and tibia were categorized as having “lower skeletal” pain. Patients who required radiotherapy to the cervical or thoracic spine, shoulder girdle/upper extremity, ribs, and skull were categorized as having “upper skeletal” bone pain. Patients who required radiotherapy to more than one area of pain were excluded from the discriminatory analysis.
      BPI assessments were completed by 312 patients. Fifty patients (16%) had missing values in one or more interference items (“normal work” missed most commonly in 23 out of 50 patients). Four patients indicated zero pain. Therefore, 258 patients were included in the analysis. Patient characteristics between those who completed the instrument in its entirety vs. those with missing item(s) were generally comparable, as shown in Table 1, with the exception that older patients were more likely to have missing responses. Examining the frequency of missing values across the seven interference items showed no difference in the pattern of missing values between patients with lower and upper skeletal pain.
      Table 1Characteristics of Patients at Time of BPI Evaluation Before Radiotherapy
      Patient/Treatment CharacteristicsStudy PatientsExcluded Patients
      Excluded patients due to missing values of BPI item(s) were significantly older (71 years vs. 67 years, t-test P=0.0280).
      No Missing BPI ValuesOne or More Missing BPI Values
      n=258n=50
      Median age at time of evaluation (IQR)67 (18)71 (16)
      Cancer type (%)
       Breast91 (35)11 (22)
       Genitourinary107 (42)23 (46)
       Lung24 (9)8 (16)
       Myeloma16 (6)3 (6)
       Others20 (8)5 (10)
      History of prior radiotherapy for bone pain (%)Yes = 53 (21)Yes = 11 (22)
      No = 205 (79)No = 39 (78)
      Median time from initial cancer diagnosis to evaluation (IQR), months39 (76)25 (89)
      Median time from bone diagnosis to evaluation (IQR), months8.1 (22)6.3 (14)
      Radiotherapy fractions (%)
       Single162 (64)28 (57)
       Multiple92 (35)21 (43)
      Radiotherapy volume (%)
       1 anatomic area196 (76)35 (70)
       >1 anatomic area62 (24)15 (30)
      Radiotherapy area (%)Total=324 areas in 258 patientsTotal=68 areas in 50 patients
      Upper skeleton (%)123 (38)24 (35)
       Cervical and thoracic spine63 (19)15 (22)
       Shoulder/humerus30 (9)4 (6)
       Ribs/sternum/clavicle/skull30 (9)5 (7)
      Lower skeleton (%)201 (65)44 (65)
       Lumbar spine+sacrum100 (31)16 (24)
       Pelvis/hip80 (24)23 (34)
       Femur/tibia/fibula21 (6)5 (7)
      Median 24-hour oral morphine consumption (IQR)30 mg equivalent daily (150 mg)29 mg equivalent daily (89 mg)
      Median Karnofsky Performance Score (IQR)
      Evaluated for 122 of 258 study patients and 21 of 50 excluded patients. Other patient characteristics were comparable.
      70 (10)60 (10)
      Mean pain score (SD)
       Worst pain6.9 (2.1)6.9 (2.6)
       Average pain5.1 (2.1)5.1 (2.4)
      IQR=interquartile range; SD=standard deviation.
      a Excluded patients due to missing values of BPI item(s) were significantly older (71 years vs. 67 years, t-test P=0.0280).
      b Evaluated for 122 of 258 study patients and 21 of 50 excluded patients. Other patient characteristics were comparable.

      Statistical Analysis

      Item Analysis and Internal Consistency

      Measures of central tendency and individual item response frequencies were summarized. Item-item correlations were examined to identify items that were redundant with other items. Internal consistency was examined with items divided into Pain and Interference subscales (two-factor model) and with items divided into Pain, Activity, and Affect subscales (three-factor model). We further examined changes in Cronbach's alpha within each subscale after the removal of individual items, noting situations where Cronbach's alpha scores increased.

      Confirmatory Factor Analysis

      Scale structure of the BPI as a single construct, two-factor and three-factor models, was examined using confirmatory factor analysis (CFA), as described by Hatcher.
      • Hatcher L.
      Developing measurement models with confirmatory factor analysis.
      Models were compared using various model fit statistics (e.g., normed Chi-square statistic, goodness of fit index, Bentler's comparative fit index). Revisions were made based on both suggested modification indices and the theoretical or practical interpretation of such changes. Covariance terms were added to the structural equations between latent factors and between error terms of indicator variable without cross-loading of indicator variable on more than one factor. Composite reliability and convergent validity were examined for the best fitting two-factor and three-factor models. Discriminant validity tests (Chi-square difference test, confidence interval test, and variance extracted test) were carried out to further evaluate highly correlated factors within the three-factor model. The resultant model of choice was applied to the predefined lower and upper skeletal groups to reveal potential differences in measured factors. Interaction was explored between skeletal grouping and mean pain intensity subgroups, stratified as mild (mean pain intensity ≤ 4), moderate (mean pain intensity > 4 and < 7), and severe (mean pain intensity ≥ 7).
      All statistical analyses were performed using SAS 9.1.3 (SAS Institute, Inc., Cary, NC). Confirmatory factor modeling was carried out using the covariance analysis of linear structural equations (PROC CALIS) procedure.

      Results

      Patient Characteristics

      Median age of the study group was 67 years, with 35% (91 out of 258) and 42% (107 out of 258) having breast and genitourinary cancers, respectively (Table 1). Sixty percent of study patients expressed severe pain, with the worst pain score being seven or more out of 10; worst pain scores in this range corresponded with expression of high interference on all items, suggesting high levels of symptoms. The median oral morphine equivalent of 30 mg/d (interquartile range 150 mg/d) indicated that most patients were already receiving strong opioid analgesics (World Health Organization ladder, step 3), and additional pain relief intervention was warranted.

      Item Analysis and Internal Consistency of Subscales

      Pain was moderately severe, with a mean of 6.9 (SD 2.1) at its worst and a mean of 5.1 (SD 2.1) on average. Interference levels were also moderate to severe, with the mean ranging from 3.9 to 7.2 across the seven items. Left-sided skewness was notable for general activity, walking, and enjoyment, with ceiling effect for normal work interference in particular. Conversely, relations with others had a right-sided skewness, with more responses in lower levels of the scale. The sleep item mapped out with a nearly uniform distribution of responses across all values from 0 to 10. The internal consistency of the subscales was high, whether items are grouped into two subscales (Pain and Interference) or three subscales (Pain, Activity, and Affect), with Cronbach's alpha ranging between 0.81 and 0.89 (Table 2). However, removing sleep interference from the subscales resulted in a noticeable increase in internal consistency. In both situations, sleep displayed low correlations with subscale totals (0.43 and 0.32), whereas other items showed consistently moderate correlations from 0.61 to 0.80.
      Table 2Summary of BPI Scores by Item and Internal Consistency Among Subscales in Study Population (n=258)
      Item Grouping
      Two SubscalesThree Subscales
      Pain Subscale (Cronbach's Alpha=0.85)Pain Subscale (Cronbach's Alpha=0.85)
      Item StatisticsCorrelation with TotalAlpha with Item DeletedCorrelation with TotalAlpha with Item Deleted
      BPI ItemsMeanSD
      Worst pain6.92.10.680.840.680.84
      Average pain5.12.10.780.740.780.74
      Pain now4.62.50.710.810.710.81
      Interference Subscale (Cronbach's Alpha=0.87)Activity Subscale (Cronbach's Alpha=0.79)
      General activity6.62.80.760.840.760.66
      Walk6.53.00.660.850.680.70
      Work7.23.00.750.840.770.65
      Sleep5.33.10.380.89↑0.260.89↑
      Affect Subscale (Cronbach's Alpha=0.80)
      Mood5.02.90.650.850.680.68
      Enjoy6.52.90.770.840.620.75
      Relations3.92.90.600.860.620.74
      High Cronbach's alpha for Pain, Interference, Activity, and Affect subscales.
      Sleep interference was least correlated with total. =when sleep was deleted, alpha increased.

      Confirmatory Factor Analysis

      Given the low correlations between sleep interference with other model subscales within this data set, we compared CFA models with and without sleep. Table 3 summarizes the corresponding model fit statistics. The “null” model of 10-item single factor showed very poor fit, which was expected given the well-established two-factor solution to the BPI. However, a two-factor default model of Pain and Interference still performed poorly with large normalized residuals (>5.0). Removing sleep from the two-factor and three-factor models improved fit considerably, with additional improvements in measures of model fit gained by allowing specific error terms to covary. Factor loadings and associated statistics for the best two-factor and three-factor models are summarized in Table 4 for comparison. Each factor (Pain, Interference, Activity, and Affect) fulfilled the minimum requirement of three indicator variables per factor. All factor loadings were highly significant, with t values between 8.7 and 17.7 (NB: minimum critical t value of 3.29 for P=0.001), supporting the convergent validity of items on their respective factor. Indicator reliability of individual items (R-squared) ranged from 0.29 (relation) to 0.93 (worst pain). Composite reliability demonstrated high internal consistency of the factors in both models, ranging from 0.76 to 1.0 (minimum acceptable level = 0.70).
      Table 3Comparison of Model Fit of Different Structural Models
      ModelModificationsGoodness of Fit Index (Adjusted)Chi-Square (df)RMSEA (Upper CL)Comparative Fit IndexNon-normed Fit Index
      1 factor: 10 itemsNone (no structure)0.56433.5 (35)0.230.720.64
      2 factors: Pain and InterferenceNone (default structure)0.76199.3 (34)0.160.880.84
      2 factors: Pain and Interference
      • Covary error terms of avgpain-pain now, mood-relation, mood-enjoy, relation-enjoy
      0.86111.2 (30)0.120.940.91
      2 factors: Pain and Interference
      • Drop sleep; covary error terms of avgpain-pain now, mood-relation, mood-enjoy, relation-enjoy
      0.9342.48 (22)0.090.980.97
      3 factors: Pain, Activity, and AffectNone (load sleep on activity)0.83152.6 (32)0.140.910.88
      3 factors: Pain, Activity, and Affect
      • Drop sleep; covary error terms of avgpain-pain now, mood-relation
      0.9437.9 (22)0.080.990.98
      Avgpain=average pain; df=degrees of freedom; CL = confidence limit.
      Modifications included covariance of error terms of paired items. Correlated error terms suggest that the two corresponding indicator variables measure something in common that is not explicitly represented in the model. Fit statistics improved when sleep was dropped from models. Goodness of fit index (adjusted for degrees of freedom) examines the ability of the model to explain the variance in the sample covariance matrix (perfect fit=1), analogous to corrected R-square. Model Chi-square (i.e., likelihood ratio Chi-square) represents the value of the statistical criterion minimized in maximum likelihood estimation (smaller value=better fit). RMSEA (root mean square error of approximation) measures the lack of fit of the model to the population covariance matrix (RMSEA ≥ 0.10 suggests poor fit). Comparative fit index (CFI) measures the improvement in the overall fit of the model compared with a model assuming that all observed variables are uncorrelated. Non-normed fit index (NNFI) correct for model complexity. CFI and NNFI above 0.9 suggest acceptable model fit.
      Table 4CFA Showing Comparable Factor Loadings and Reliability of Both 2-Factor and 3-Factor Models (Sleep Interference Excluded From Models)
      Variables2-Factor ModelVariables3-Factor Model
      Factor LoadingR-Squaredt-StatisticFactor LoadingR-Squaredt-Statistic
      Worst pain0.960.9314.0Worst pain0.950.9114.3
      Average pain0.710.5410.7Average pain0.710.5010.9
      Pain now0.610.409.3Pain now0.620.389.5
      Composite Reliability Pain Factor 0.81Composite Reliability Pain Factor 0.81
      General activity0.890.8017.7General activity0.890.7917.6
      Walking0.810.6615.3Walking0.810.6515.3
      Work0.880.7917.3Work0.890.7817.4
      Composite Reliability Activity Factor 0.90
      Mood0.530.308.7Mood0.650.4310.9
      Enjoyment0.740.5713.4Enjoyment0.880.7815.8
      Relations0.530.298.7Relations0.600.369.8
      Composite Reliability Interference Factor 1.0Composite Reliability Affect Factor 0.76
      Correlation Pain-Interference0.54Correlation Pain-Activity0.52
      Correlation Pain-Affect0.55
      Correlation Activity-Affect0.83
      For the three-factor model, moderate but near-identical correlations of Pain-Activity (correlation = 0.52) and Pain-Affect (correlation = 0.55) were observed, with high correlation between Activity and Affect (correlation = 0.83), suggesting that Activity and Affect could be representative of the same latent variable. Tests of discriminant validity provided mixed support for Activity and Affect: the Chi-square difference test (Chi-square difference 28.4, df=1, P<0.001) and the confidence interval test (correlation 0.83, 95% confidence interval 0.76–0.90) were significant, but the variance extracted test failed to confirm discriminant validity because the variance extracted estimate for Affect (0.52) was lower than the square of correlation between Activity and Affect (0.69), even though the variance extracted estimate for Activity was high (0.74).
      In summary, both the two-factor (Pain and Interference) and the three-factor models (Pain, Activity, and Affect) demonstrated reasonably high levels of internal consistency, composite reliability, and convergent validity. Overall, the three-factor model was preferred because fewer covariance terms were needed to fit the model, and those three factors represented three of the four IMMPACT core domains, which could be useful in differentiating patients with different sources of pain. The final three-factor model (excluding sleep) with standardized factor loadings and correlations is illustrated in Fig. 1.
      Figure thumbnail gr1
      Fig. 1Final measurement model of BPI evaluations of 258 bone metastases patients. Model showing 3 distinct factors: Pain, Activity interference, and Affect interference with standardized factor loading of each item onto respective factor. E = residual for each item; C = correlation estimate. Model's adjusted goodness of fit = 0.94.

      Pattern of Pain, Activity, and Affect Interference in Lower vs. Upper Skeletal Pain

      Table 5 presents mean scores of BPI measurements according to three factors of Pain, Activity interference, and Affect interference of patients treated for either upper skeletal or lower skeletal bone metastases (patients requiring radiotherapy to more than one volume were excluded). Even though the mean Pain scores were nearly identical between the lower and upper skeletal groups (5.7 vs. 5.6, respectively), the mean Activity score was significantly worse in the lower skeletal group (7.4 vs. 5.7, P<0.0001), thus demonstrating a clinical and significant detriment in activity function as a result of skeletal metastases in the lower skeleton. Further examination of mean Activity score in Pain subgroups, stratified as mild, moderate, and severe categories, demonstrated greater impact of lower skeletal pain on Activity among patients expressing mild to moderate pain (Table 6). Patients already experiencing severe pain suffered severe interference of function, regardless of site of skeletal involvement.
      Table 5Summary of Subscale Mean Scores in Patients with Upper vs. Lower Skeletal Pain
      Mean score represents average of item scores within each subscale (e.g., mean Pain score=average of worst pain, average pain, and pain now).
      SubscalesUpper Skeletal Pain (n=72)Lower Skeletal Pain (n=124)
      MeanSDMeanSD
      Mean Pain
      Mean score represents average of item scores within each subscale (e.g., mean Pain score=average of worst pain, average pain, and pain now).
      5.72.05.62.0
      Mean Activity
      Mean activity of lower skeletal pain group significantly higher than upper skeletal pain group (t=4.2, P<0.0001). Mean activity difference of 1.7 (95% confidence limits 0.9–2.5) and effect size of 0.68 suggest clinically meaningful difference between these two groups.
      5.72.97.42.3
      Mean Affect5.02.65.12.4
      a Mean score represents average of item scores within each subscale (e.g., mean Pain score=average of worst pain, average pain, and pain now).
      b Mean activity of lower skeletal pain group significantly higher than upper skeletal pain group (t=4.2, P<0.0001). Mean activity difference of 1.7 (95% confidence limits 0.9–2.5) and effect size of 0.68 suggest clinically meaningful difference between these two groups.
      Table 6Mean Activity Scores in Upper and Lower Skeletal Patient Groups Stratified by Pain Subgroups (Mild, Moderate, and Severe Mean Pain Scores)
      Pain Subgroups
      Test of heterogeneity of effect on mean Activity scores between pain subgroups and upper/lower skeletal groups (interaction term) shows statistical significance (P=0.026, general linear model), suggesting that effect of pain on Activity interference is modified by location of skeletal involvement.
      Upper Skeletal Pain (n=72)Lower Skeletal Pain (n=124)
      nMean ActivitySDnMean ActivitySD
      Mild (mean Pain ≤ 4)
      Significantly higher mean Activity interference in lower skeletal group among patients expressing mild pain, mean difference 2.6 (95% confidence limits 1.1–4.2, t=3.4, P=0.0016), and moderate pain, mean difference 1.9, (95% confidence limits 0.9–2.9, t=4.1, P=0.0004), but not among patients with severe pain, mean difference 0.6 (95% confidence limits −0.5 to 1.6, t=1.1, P=0.29).
      173.42.1276.02.8
      Moderate (mean Pain 4–7)
      Significantly higher mean Activity interference in lower skeletal group among patients expressing mild pain, mean difference 2.6 (95% confidence limits 1.1–4.2, t=3.4, P=0.0016), and moderate pain, mean difference 1.9, (95% confidence limits 0.9–2.9, t=4.1, P=0.0004), but not among patients with severe pain, mean difference 0.6 (95% confidence limits −0.5 to 1.6, t=1.1, P=0.29).
      335.62.8617.52.2
      Severe (mean Pain ≥ 7)
      Significantly higher mean Activity interference in lower skeletal group among patients expressing mild pain, mean difference 2.6 (95% confidence limits 1.1–4.2, t=3.4, P=0.0016), and moderate pain, mean difference 1.9, (95% confidence limits 0.9–2.9, t=4.1, P=0.0004), but not among patients with severe pain, mean difference 0.6 (95% confidence limits −0.5 to 1.6, t=1.1, P=0.29).
      227.52.4368.11.7
      a Test of heterogeneity of effect on mean Activity scores between pain subgroups and upper/lower skeletal groups (interaction term) shows statistical significance (P=0.026, general linear model), suggesting that effect of pain on Activity interference is modified by location of skeletal involvement.
      b Significantly higher mean Activity interference in lower skeletal group among patients expressing mild pain, mean difference 2.6 (95% confidence limits 1.1–4.2, t=3.4, P=0.0016), and moderate pain, mean difference 1.9, (95% confidence limits 0.9–2.9, t=4.1, P=0.0004), but not among patients with severe pain, mean difference 0.6 (95% confidence limits −0.5 to 1.6, t=1.1, P=0.29).

      Discussion

      The goal of the present study was to first validate the psychometric properties of the BPI in patients with clinically significant metastatic bone pain requiring palliative radiotherapy and then to evaluate whether BPI subscales were able to discriminate clinically important groups of bone metastases patients, specifically with relation to the site of the pain (lower vs. upper skeletal pain). We verified that the psychometric properties of the BPI instrument were robust in our population of advanced cancer pain patients referred for palliative radiotherapy. Either a two-factor (Pain and Interference) or a three-factor (Pain, physical interference or Activity, psychosocial interference or Affect) confirmatory factor model demonstrated adequate model fit. We also confirmed our clinical suspicion that interference with activity was significantly worse in patients with metastatic bone pain to the lower skeletal areas compared with those with upper skeletal pain alone. This clinically substantial difference in activity interference was most pronounced in patients indicating only mild to moderate pain intensity.
      The main limitation of this validation study is the absence of data from other validated tools, which would have allowed multitrait-multimethod analysis. Although our study has provided initial evidence of differences in pain subscales in relation to site of bone pain, further evaluation and validation with other musculoskeletal function scales, such as the Western Ontario and McMaster University Osteoarthritis Index (WOMAC) WOMAC
      • Bellamy N.
      • Buchanan W.W.
      • Goldsmith C.H.
      • Campbell J.
      • Stitt L.W.
      Validation study of WOMAC: a health status instrument for measuring clinically important patient relevant outcomes to antirheumatic drug therapy in patients with osteoarthritis of the hip or knee.
      or the Toronto Extremity Salvage Score,
      • Davis A.M.
      • Wright J.G.
      • Williams J.I.
      • et al.
      Development of a measure of physical function for patients with bone and soft tissue sarcoma.
      could help confirm our findings. Others have validated BPI results against the Short-Form 36 Health Survey (SF36) or the European Organization for Research and Treatment of Cancer quality of life 30-item core questionnaire (EORTC QLQC-30) questionnaires,
      • Radbruch L.
      • Loick G.
      • Kiencke P.
      • et al.
      Validation of the German version of the Brief Pain Inventory.
      • Klepstad P.
      • Loge J.H.
      • Borchgrevink P.C.
      • et al.
      The Norwegian Brief Pain Inventory questionnaire: translation and validation in cancer pain patients.
      but none attempted to characterize pain by comparing BPI measurements with instruments developed for the musculoskeletal system. Considering the intermittent nature of metastatic bone pain, further studies also may aim at exploring the possible interaction between breakthrough pain and anatomic sites on functional interference, which may lead to risk stratification among the heterogeneous population of bone metastases patients.
      Despite its brevity, one in six referred patients in our study population was unable to complete the BPI in its entirety. The most commonly missed item was normal work, which contained the qualifying phrase “both work outside the home and housework.” Elderly patients, particularly those requiring palliative cancer therapies, might have felt that they were not required to perform housework or employed work, resulting in an absence of rating on this item. Further consideration of item wording and sensibility might improve the performance of this item or identify potential substitution. Because we have excluded this small group of elderly patients from factor analysis, we caution against generalizing our findings to this subpopulation.
      Our analysis of pain experience in patients with lower vs. upper skeletal pain did not take into account potential confounders, such as age, disease severity, and performance status, because some variables were not recorded consistently. With respect to test-retest reliability and longitudinal construct validity, our practice was not designed to consistently require repeated measurements—which limited our ability to examine the BPI as an evaluative tool to detect change—or to show how relationships among symptom clusters might change over time, as suggested by Hadi et al.
      • Hadi S.
      • Fan G.
      • Hird A.E.
      • et al.
      Symptom clusters in patients with cancer with metastatic bone pain.
      Recently completed and ongoing clinical trials of palliative radiotherapy for bone pain using the BPI as a pain instrument are expected to provide opportunities to explore such issues in this setting.
      • Hartsell W.F.
      • Scott C.B.
      • Bruner D.W.
      • et al.
      Randomized trial of short- versus long-course radiotherapy for palliation of painful bone metastases.
      • Chow E.
      • Hoskin P.J.
      • Wu J.
      • et al.
      A phase III international randomised trial comparing single with multiple fractions for re-irradiation of painful bone metastases: National Cancer Institute of Canada Clinical Trials Group (NCIC CTG) SC 20.
      As far as we are aware, our study is the first validation study of the BPI in patients with skeletal metastases. Exploring patterns of pain and its interference revealed a subpopulation of patients whose relatively mild pain intensity could conceal clinically significant functional impairment. Attention to activity function is critical when managing patients with lower body skeletal metastases. Other studies have distinguished subgroups according to mild, moderate, and severe pain intensity
      • Serlin R.C.
      • Mendoza T.R.
      • Nakamura Y.
      • Edwards K.R.
      • Cleeland C.S.
      When is cancer pain mild, moderate or severe? Grading pain severity by its interference with function.
      and observed differences between cancer and non-cancer patients in scoring interference.
      • Holen J.C.
      • Lydersen S.
      • Klepstad P.
      • Loge J.H.
      • Kaasa S.
      The Brief Pain Inventory: pain's interference with functions is different in cancer pain compared with noncancer chronic pain.
      In our practice, patients are typically referred for palliative radiotherapy when pain is moderately severe or is unresponsive to opioids. But as lower skeletal metastases appear to confer greater interference with function, such patients may benefit from routine monitoring or early referral for proactive management before pain becomes severe. Apart from pain relief, increased bone mineral density in lytic metastatic lesions has been demonstrated after focal palliative radiotherapy.
      • Koswig S.
      • Budach V.
      This provides a reasonable conjecture toward a strategy of fast-tracking patients for radiotherapy or, alternatively, orthopedic evaluation based on the site of skeletal involvement (i.e., lumbosacral spine, pelvis/hip, femur), focusing on its interference with activity more than its pain intensity.
      In this highly symptomatic population of patients, we were surprised to find that sleep interference was not contributory to either the Activity or Affect subscale. Possible explanations include adaptability of patients to interference of sleep by pain (e.g., use of physical measures to splint painful body parts, increasing time in bed to make up for disrupted sleep), variable interpretation of item wording, and variable effects of opioid analgesics on either sedating or disrupting normal sleep patterns. The lack of internal consistency of sleep interference in either Activity or Affect can be discerned from previous studies where sleep was included with Activity in one study,
      • Cleeland C.S.
      • Nakamura Y.
      • Mendoza T.R.
      • et al.
      Dimensions of the impact of cancer pain in a four country sample: new information from multidimensional scaling.
      with Affect in another,
      • Klepstad P.
      • Loge J.H.
      • Borchgrevink P.C.
      • et al.
      The Norwegian Brief Pain Inventory questionnaire: translation and validation in cancer pain patients.
      and excluded in another.
      • Saxena A.
      • Mendoza T.
      • Cleeland C.S.
      The assessment of cancer pain in north India: the validation of the Hindi Brief Pain Inventory—BPI-H.
      Sleep disorders are prevalent in cancer patients, are correlated with pain, and are typically multifactorial.
      • McMillan S.C.
      • Tofthagen C.
      • Morgan M.A.
      Relationships among pain, sleep disturbances, and depressive symptoms in outpatients from a comprehensive cancer center.
      Clearly, further study of the interrelatedness of pain and sleep disorders is needed to understand this complex area.
      In summary, the BPI is a valid and reliable tool for pain measurement in patients with bone metastases. We have demonstrated its discriminant properties in a cohort of patients referred for palliative radiotherapy. Sleep interference did not add to the pain construct, and elderly patients had difficulty responding to interference items, particularly regarding housework or work outside home. CFA of the BPI items identified comparable two-factor and three-factor models, but the latter is recommended for its relatively simple covariance structure. Patients with metastatic lesions in the lower body experience significant interference on activity even when pain is mild to moderate and may require particular attention because of higher risk of functional impairment.

      Acknowledgments

      The authors thank Dr. Charles Cleeland for his permission to use the BPI.

      References

        • van den Beuken-van Everdingen M.H.
        • de Rijke J.M.
        • Kessels A.G.
        • et al.
        Prevalence of pain in patients with cancer: a systematic review of the past 40 years.
        Ann Oncol. 2007; 18: 1437-1449
        • Goldberg G.R.
        • Morrison R.S.
        Pain management in hospitalized cancer patients: a systematic review.
        J Clin Oncol. 2007; 25: 1792-1801
        • Cleeland C.S.
        • Gonin R.
        • Hatfield A.K.
        • et al.
        Pain and its treatment in outpatients with metastatic cancer.
        N Engl J Med. 1994; 330: 592-596
        • Cleeland C.S.
        The measurement of pain from metastatic bone disease: capturing the patient's experience.
        Clin Cancer Res. 2006; 12: 6236s-6242s
        • Mercadante S.
        • Fulfaro F.
        Management of painful bone metastases.
        Curr Opin Oncol. 2007; 19: 308-314
        • Dworkin R.H.
        • Turk D.C.
        • Wyrwich K.W.
        • et al.
        Interpreting the clinical importance of treatment outcomes in chronic pain clinical trials: IMMPACT recommendations.
        J Pain. 2008; 9: 105-121
        • Caraceni A.
        • Mendoza T.R.
        • Mencaglia E.
        • et al.
        A validation study of an Italian version of the Brief Pain Inventory (Breve Questionario per la Valutazione del Dolore).
        Pain. 1996; 65: 87-92
        • Radbruch L.
        • Loick G.
        • Kiencke P.
        • et al.
        Validation of the German version of the Brief Pain Inventory.
        J Pain Symptom Manage. 1999; 18: 180-187
        • Wang X.S.
        • Mendoza T.R.
        • Gao S.Z.
        • Cleeland C.S.
        The Chinese version of the Brief Pain Inventory (BPI-C): its development and use in a study of cancer pain.
        Pain. 1996; 67: 407-416
        • Uki J.
        • Mendoza T.
        • Cleeland C.S.
        • Nakamura Y.
        • Takeda F.
        A brief cancer pain assessment tool in Japanese: the utility of the Japanese Brief Pain Inventory—BPI-J.
        J Pain Symptom Manage. 1998; 16: 364-373
        • Saxena A.
        • Mendoza T.
        • Cleeland C.S.
        The assessment of cancer pain in north India: the validation of the Hindi Brief Pain Inventory—BPI-H.
        J Pain Symptom Manage. 1999; 17: 27-41
        • Klepstad P.
        • Loge J.H.
        • Borchgrevink P.C.
        • et al.
        The Norwegian Brief Pain Inventory questionnaire: translation and validation in cancer pain patients.
        J Pain Symptom Manage. 2002; 24: 517-525
        • Kalyadina S.A.
        • Ionova T.I.
        • Ivanova M.O.
        • et al.
        Russian Brief Pain Inventory: validation and application in cancer pain.
        J Pain Symptom Manage. 2008; 35: 95-102
        • Cleeland C.S.
        • Nakamura Y.
        • Mendoza T.R.
        • et al.
        Dimensions of the impact of cancer pain in a four country sample: new information from multidimensional scaling.
        Pain. 1996; 67: 267-273
        • Hagen N.A.
        • Biondo P.
        • Stiles C.
        Assessment and management of breakthrough pain in cancer patients: current approaches and emerging research.
        Curr Pain Headache Rep. 2008; 12: 241-248
        • Fainsinger R.L.
        • Nekolaichuk C.L.
        A “TNM” classification system for cancer pain: the Edmonton Classification System for Cancer Pain (ECS-CP).
        Support Care Cancer. 2008; 16: 547-555
        • Wu J.S.
        • Monk G.
        • Clark T.
        • et al.
        Palliative radiotherapy improves pain and reduces functional interference in patients with painful bone metastases: a quality assurance study.
        Clin Oncol (R Coll Radiol). 2006; 18: 539-544
        • Daut R.L.
        • Cleeland C.S.
        • Flanery R.C.
        Development of the Wisconsin Brief Pain Questionnaire to assess pain in cancer and other diseases.
        Pain. 1983; 17: 197-210
        • Chow E.
        • Wu J.S.
        • Hoskin P.
        • et al.
        International consensus on palliative radiotherapy endpoints for future clinical trials in bone metastases.
        Radiother Oncol. 2002; 64: 275-280
        • Hatcher L.
        Developing measurement models with confirmatory factor analysis.
        in: A step-by-step approach to using SAS for factor analysis and structural equation modeling. SAS Publishing, Cary, NC2007: 249-340
        • Bellamy N.
        • Buchanan W.W.
        • Goldsmith C.H.
        • Campbell J.
        • Stitt L.W.
        Validation study of WOMAC: a health status instrument for measuring clinically important patient relevant outcomes to antirheumatic drug therapy in patients with osteoarthritis of the hip or knee.
        J Rheumatol. 1988; 15: 1833-1840
        • Davis A.M.
        • Wright J.G.
        • Williams J.I.
        • et al.
        Development of a measure of physical function for patients with bone and soft tissue sarcoma.
        Qual Life Res. 1996; 5: 508-516
        • Hadi S.
        • Fan G.
        • Hird A.E.
        • et al.
        Symptom clusters in patients with cancer with metastatic bone pain.
        J Palliat Med. 2008; 11: 591-600
        • Hartsell W.F.
        • Scott C.B.
        • Bruner D.W.
        • et al.
        Randomized trial of short- versus long-course radiotherapy for palliation of painful bone metastases.
        J Natl Cancer Inst. 2005; 97: 798-804
        • Chow E.
        • Hoskin P.J.
        • Wu J.
        • et al.
        A phase III international randomised trial comparing single with multiple fractions for re-irradiation of painful bone metastases: National Cancer Institute of Canada Clinical Trials Group (NCIC CTG) SC 20.
        Clin Oncol (R Coll Radiol). 2006; 18: 125-128
        • Serlin R.C.
        • Mendoza T.R.
        • Nakamura Y.
        • Edwards K.R.
        • Cleeland C.S.
        When is cancer pain mild, moderate or severe? Grading pain severity by its interference with function.
        Pain. 1995; 61: 277-284
        • Holen J.C.
        • Lydersen S.
        • Klepstad P.
        • Loge J.H.
        • Kaasa S.
        The Brief Pain Inventory: pain's interference with functions is different in cancer pain compared with noncancer chronic pain.
        Clin J Pain. 2008; 24: 219-225
        • Koswig S.
        • Budach V.
        Strahlenther Onkol. 1999; 175: 500-508
        • McMillan S.C.
        • Tofthagen C.
        • Morgan M.A.
        Relationships among pain, sleep disturbances, and depressive symptoms in outpatients from a comprehensive cancer center.
        Oncol Nurs Forum. 2008; 35: 603-611