364 Journal of Pain and Symptom Management Vol. 16 No. 6 December 1998 Original Article A Brief Cancer Pain Assessment Tool in Japanese: The Utility of the Japanese Brief Pain Inventory BPI-J Jiro Uki, MD, Tito Mendoza, MS, Charles S. Cleeland, PhD, Yoshio Nakamura, PhD, and Fumikazu Takeda, MD Saitama Cancer Center and World Health Organization Collaborating Center for Cancer Pain Relief and Quality of Life, Ina, Japan (J.U., F.T.); Pain Research Group, University of Texas M. D. Anderson Cancer Center, Houston, TX, USA (T.M., C.S.C.); and Talaria, Inc, Seattle, WA, USA (Y.N.) Abstract Despite recognition that pain management is an important component in the treatment of Japanese cancer patients, progress in this area has been slowed by the lack of an appropriate measure of pain. In a prospective, single-institution study, a Japanese translation of the Brief Pain Inventory (BPI-J) was administered to 121 patients to assess the intensity and impact of cancer-related pain. After an analysis of the instrument s reliability and validity, this study tested the utility of the new measure in an analysis of the adequacy of analgesic prescription. Results were compared with predictive models from studies using non-japanese patient groups. Factor analysis of the BPI items resulted in two factors, pain severity and pain interference, showing consistency with other language versions of the tool. Coefficient alphas of greater than 0.80 for the items comprising these two subscales indicate a reliable self-report pain instrument. After establishing the validity and reliability of the BPI-J, we examined possible predictors of inadequate pain management in these Japanese cancer patients. Similar to studies done in other countries, women patients and those whose pain severity was underestimated by their physician were more likely to be undermanaged for pain. The results of this study support the utility of the Japanese BPI for studies of the epidemiology of cancer pain in Japan, as well as for the assessment of pain treatment outcome in Japanese-speaking patients. J Pain Symptom Manage 1998;16:364 373. U.S. Cancer Pain Relief Committee, 1998. Key Words Pain management, pain assessment, supportive care Introduction The Brief Pain Inventory (BPI) is a quickly administered assessment tool for capturing both the intensity of a patient s cancer-related pain Address reprint requests to: Charles S. Cleeland, Ph.D., Pain Research Group, 1100 Holcombe Blvd, Box 221, Houston, TX 77030. Accepted for publication: April 13, 1998. as well as the amount of interference in a patient s life caused by pain. First developed in English, the BPI s simple format and its focus on a limited number of relatively universal functions has made it easy to translate. Several studies have provided evidence that patients who are from different cultures and speak different languages rate the items in a similar fashion. 1,2 English and Spanish versions of the BPI have been used in multicenter studies of cancer U.S. Cancer Pain Relief Committee, 1998 0885-3924/98/$ see front matter Published by Elsevier Science, New York, New York PII S0885-3924(98)00098-0
Vol. 16 No. 6 December 1998 Japanese Brief Pain Inventory 365 pain and its treatment in the United States, 3 and a French version has been used in a similar multicenter study in France. 4 Originally designed for cancer patients, whose limited endurance mandated the choice of a very few relevant items, the BPI and other language versions have also been used in multicenter studies of AIDS-related pain and its treatment in the United States. 5 An estimate derived from the BPI, the Pain Management Index, 3 has been used as an estimate of the adequacy of physicians cancer pain management, comparing the severity of pain to the use of accepted guidelines for analgesic prescribing. 6 If the BPI continues to demonstrate consistent measurement characteristics across different languages and with diverse cultural groups, the possibility of conducting multinational studies of the epidemiology and treatment of pain is strengthened, as is the possibility of conducting multinational analgesic clinical trials. The easiest translation process is from one European-based language (English) to others (French, Spanish). A more demanding test is to examine the performance of the BPI with Asian languages and cultures. An early study 7 provided evidence that a Vietnamese version of the BPI yielded similar measurement dimensions to the English version, and a second study, reported in Serlin et al., 1 found that a Filipino version also had similar measurement characteristics. Perhaps even a more challenging task is the translation of a self-report measurement instrument into character-based written languages, such as Chinese or Japanese. We have recently reported the development and validation of a Chinese character version of the BPI, the BPI-C, and have offered evidence that the BPI-C yields similar dimensions to other language versions. 8 Chinese cancer patients, rating their pain and its impact on the BPI-C, reported a similar relationship of increasing severity of pain with associated increases in functional impairment. 1 We also found that the Pain Management Index, derived from the BPI, accurately reflected the current status of cancer pain management in Chinese hospitals. 8 Previous attempts to translate pain assessment instruments into Japanese have been difficult because of language and cultural differences. 9 We now report the development and validation of a Japanese version of the BPI, the BPI-J. Although the validation study was confined to a single cancer center, we also examined if some of the predictive relationships among disease, demographic variables, and the BPI, which were found in other language and national samples, could be found in this sample. Finding such relationships would present additional evidence for the utility of the instrument for both research and clinical applications. Methods Subjects The validation sample consisted of 121 patients from the Saitama Cancer Center who had a pathological diagnosis of cancer and also experienced pain. The mean age was 56 (range 19 to 82), and 45% of the sample were female. Eighty-eight percent had either recurrent or metastatic disease, and 86% were receiving active pain treatment. All the patients had a definitive pain syndrome due to their neoplastic disease rather than as a result of antineoplastic therapy or noncancer causes, as determined by their attending oncologist. The top three primary diagnoses were: lung cancer (22%), breast cancer (13%), and stomach cancer (12%). Pain Measurement Development of the BPI-J The Japanese version of the Brief Pain Inventory (BPI-J) was translated by the first author (J.U.) and back translated by a professional bilingual translator unfamiliar with either the English or Japanese versions of the BPI to assure accuracy of rendition. The English and Japanese versions of the BPI (Short Form) are presented as Appendix I and II (see pages 370 373). The BPI asks patients to rate their pain at its worst for the last 24 hours on a 0 to 10 scale. They also rate their pain at its least, its average, and at the time of the study on three additional 0 to 10 scales. Each scale is presented as a horizontal row of equidistant numbers from 0 to 10, and is bounded by the words no pain at the 0 end and pain as bad as you can imagine at the other. Using the same type of 0 to 10 scales, patients are also asked to rate separately how their pain interferes with several life domains, including enjoyment of life, activity, walking, mood, sleep, work, and relations with others. These scales are bounded by does not interfere at the 0 end and interferes completely at the other end. The patients are also
366 Uki et al. Vol. 16 No. 6 December 1998 asked to estimate the percentage of pain relief they were receiving from their pain treatment, to locate areas of pain on a human figure, and to estimate the cause of their pain (cancer, cancer treatment, or noncancer). Statistical Analysis Factor analyses of data based on other language versions of the BPI consistently demonstrate that two interpretable factors are produced, one consisting of the pain severity items and the second consisting of the pain interference items. Based on these studies, the BPI can be thought of as consisting of a pain severity subscale and a pain interference subscale. In assessing the reliability of the BPI-J, coefficient alphas were computed for items comprising both the pain severity and pain interference subscales. These reliability measures were computed before testing the goodness of fit of the two-factor model to the BPI-J. To establish construct validity, principal axis factor analysis with nonorthogonal (oblimin) rotation, allowing the factors to be correlated, was then utilized in extracting factors. 10,11 Both two-factor and three-factor solutions were fitted to the data. Competing factor solutions were evaluated in terms of the distribution of factor loadings over two versus three factors. The appropriate factor solution was then chosen based on the criteria of interpretability and fit. Predictors of Inadequate Pain Management Several studies examining the adequacy of analgesic management of cancer pain have utilized a conservative index of treatment adequacy to assess whether or not analgesics were prescribed to treat pain according to the World Health Organization s guidelines for the management of pain in cancer. 6 This index, called the Pain Management Index (PMI), considers pain management appropriate when there is congruence between the patient s reported level of pain and the potency of the prescribed analgesic drug. The PMI is derived from the pain worst score of the BPI. It compares the most potent analgesic prescribed for a patient relative to the level of that patient s reported pain. For a more thorough description of this index, see Cleeland et al. 3 In general, negative PMI scores are considered to be one indicator of inadequate prescription of analgesic drugs, and scores of 0 or greater are considered to be a conservative indicator of acceptable analgesic prescribing. Multiple regression was used to examine possible predictors of inadequate analgesic prescribing, as defined by the PMI. Previous studies 3,5 have identified several predictors, including age, gender, physician s estimate of pain severity, performance status, and ethnic heritage. Model fit was assessed using change in R 2. Stepwise, forward, and backward selection procedures were all used to compare solutions. Results Table 1 presents descriptive statistics for the BPI items from the Japanese sample. The magnitude of pain severity and pain interference items from this single hospital sample were generally lower than those seen in other patient samples we have studied. 1 4 Item Reliability To assess the reliability of the BPI-J, we separately calculated coefficient alpha for the severity scale (four pain severity items) and for the pain interference scale (seven interference items). Both coefficients were 0.81 (compared with 1.0 as perfect reliability), demonstrating the relatively good internal consistencies of the scales. 11 If an item is deleted, coefficient alphas lie between 0.78 to 0.80 for the interference scale, and between 0.70 to 0.80 for the severity scale. These alpha values for the scale when an item is deleted are comparable to the overall values for each of the two scales. This suggests that each of the items contributes similarly to Table 1 Descriptive Statistics for the BPI Items a Variables Mean Standard deviation Severity items Worst 4.89 2.60 Average 3.82 1.92 Now 2.88 2.23 Least 1.93 1.59 Interference items Mood 4.31 2.76 Enjoy 3.95 3.15 Activity 3.73 2.75 Sleep 3.62 2.91 Walk 3.34 3.04 Relate 3.19 2.81 Work 2.97 2.96 a The BPI-J data show that patients rated interference with mood as the highest, closely followed by enjoyment of life, activity, sleep, walk, relationships with others, and work in order of decreasing magnitude.
Vol. 16 No. 6 December 1998 Japanese Brief Pain Inventory 367 the underlying construct it is intended to measure. The reliability of the scales in the BPI-J compares favorably with that of the scales of the BPI for other countries. 1 Table 2 Factor Loadings of a Two-factor Solution Using Principal Axis Factoring with Nonorthogonal Rotation a Factor 1 Factor 2 Interference items Activity 0.77 0.14 Enjoy 0.62 0.01 Mood 0.62 0.13 Work 0.61 0.09 Relate 0.61 0.07 Walk 0.53 0.20 Sleep 0.47 0.13 Severity items Average 0.09 0.90 Now 0.01 0.83 Worst 0.06 0.58 Least 0.13 0.55 a Items are grouped according to the magnitude of the factor loading. Factor 1, the Interference Factor, is composed of items measuring how pain interferes with functional activity; Factor 2, the Severity Factor, includes items assessing pain severity. Bold numbers highlight the factor to which the items belong. Establishing Construct Validity To compare the structural similarities of the BPI-J to other language versions of the BPI, confirmatory factor analysis was used to examine construct validity. We used principal axis factor solution with a direct nonorthogonal rotation. 12 Confirmatory factor analysis involves testing a known phenomenon, as dictated by theory or previous knowledge, with a set of new data. 10 In this case, we wanted to determine whether the underlying structure of the BPI-J was the same as the structure of the translations of the BPI developed in other languages. Factor analyses using other language versions of the BPI show that the items are distributed across two factors; the items from the pain interference scale form one factor, and the items from the pain severity scale form the other. This two-factor solution (as opposed to a one- or three-factor solution) was deemed appropriate by examining the magnitude and rate of change in the eigenvalues. In this case, the eigenvalues were 4.30, 1.7, 0.98, and 0.74 for 1, 2, 3, and 4 factors, respectively. Based on the rule that meaningful factors should be associated with eigenvalues greater than 1.0 and a marginal change occurs after two factors (scree test), the two-factor solution is appropriate. 10,12 The factor loadings for the two-factor solution are presented in Table 2. Finally, to test the adequacy of this two-factor solution for the BPI-J, the differences between the reproduced correlations based on the twofactor pattern solution and the observed correlations were examined. Harman 10 provided a criterion for judging the adequacy of the solution: the standard deviation of the residuals should be less than or approximately equal to the standard error of a correlation coefficient defined as the reciprocal of the square root of the sample size. In the case of the two-factor solution, the standard deviation of the residuals at 0.02 is less than the reciprocal of the square root of 121, which is 0.09. As a further evaluation of the utility of the BPI-J, we investigated a predictive model for the values of the PMI for patients in this study. Forty-two percent and 51% of samples of cancer patients from the U.S. and France, respectively, were considered as not receiving adequate analgesia using the PMI. 3,4 In contrast, only 27% of the cancer patients in this singlesite Japanese sample had a negative PMI. Although previous studies conducted in the U.S. and France indicated that racial heritage, stage of disease, and presence of metastasis are possible predictors, only Eastern Cooperative Oncology Group (ECOG) performance status, discrepancy between a physician s and a patient s ratings of pain, gender, and age were collected for this study and used in this analysis. In some analyses of previous data sets, we have found a significant gender by discrepancy interaction. Thus, possible interactions were also investigated for this Japanese sample. Finally, stepwise, forward, and backward selection procedures using a set of predictors described above resulted in the same model. Multiple regression analysis indicated that both gender and physician underestimation of their patient s pain significantly predict the adequacy of physician analgesic prescribing. Table 3 shows both the unstandardized and standardized regression coefficients for discrepancy and gender. The 95% confidence intervals for the unstandardized regression coefficients are also given. Patients whose pain was underestimated by their physicians were more likely to be inadequately managed. For every unit on the numeric rating scale that the physicians and patients disagree, there is about 0.16
368 Uki et al. Vol. 16 No. 6 December 1998 Dependent variable: PMI ( 3 worst pain mgt, 3 best pain mgt) Table 3 Predictors of Inadequate Pain Management in Japan a Unstandardized regression coefficient (95% CI) Standardized regression coefficient P value Constant 0.08 Gender (0 female, 1 male) 0.60 (0.15, 1.04) 0.22 0.01 Discrepancy ( 10 underestimate, 10 overestimate) 0.16 (0.10, 0.23) 0.41 0.01 a Patients whose physicians underestimated their pain tend to be inadequately treated for pain. Females are also likely to be undertreated for pain. unit change in treatment adequacy across gender. Female patients are also likely to be undermanaged for their pain by approximately half a point on the index, provided discrepancy is kept constant. Together, these two predictors accounted for approximately 23% of the variability in the adequacy of pain treatment. Discussion Age-adjusted death rates for cancer in Japan are similar to other developed countries, such as the United States and Canada. One of every four deaths in Japan is due to cancer, and both the public and health professionals are becoming more aware of the importance of cancer pain relief. 13 The national government has published manuals for care of the terminally ill, and information on cancer pain relief has been disseminated through medical meetings, publications, and mass communication. Narcotics regulations have been amended in order to improve the accessibility of opioids, especially morphine, to cancer patients with pain. Although these activities have resulted in a 35- fold increase in the medical consumption of morphine between 1979 and 1992, the annual consumption per capita is still much smaller than that in other developed countries. 13 Inadequate assessment of cancer pain, either recognized by health professionals themselves 14 or as indicated by the underestimation of pain in survey studies, 3,5 has been strongly linked to the underprescription of analgesics for the management of cancer pain. The need for accurate assessment has been recognized as a barrier to cancer pain treatment in Japan, but previous attempts to translate pain assessment tools have been frustrated by differences in cultural background and language. 9 Although, by comparison with studies in other countries, the patients in this sample from a major cancer center in Japan were generally adequately treated, physician underestimation of pain was the major predictor of underprescription with morphine or other strong opioids. As has been found in studies in the United States, women were also at greater risk for underprescription of analgesics, a finding that may also be related to inadequate assessment. 3 The index of inadequate analgesic prescription was very low at this single institution, compared with similar studies in other countries. While this may be related to the lower mean pain severity ratings obtained from this sample, it should also be noted that this hospital has been the national demonstration site for the practice of the World Health Organization s cancer pain analgesic ladder, 6 and the level of pain management there may be unrepresentative of other hospitals in Japan. This study provides evidence that, with the BPI-J, Japanese cancer patients rate their pain along the same two dimensions (pain severity and pain interference) used by patients responding to other, previously validated language versions of the BPI. Factor analysis indicates that the BPI-J is composed of the severity and interference factors seen in other language versions, and that these scales are internally consistent. Finally, a convincing link is demonstrated in this single sample of Japanese patients between some previously identified variables associated with undertreatment of cancer pain and a BPI-derived estimate of inadequate analgesic prescription, providing additional evidence of the utility of the BPI-J. The BPI-J may prove useful for clinical assessment as well as Japanese analgesic trials and studies of the epidemiology of pain in Japan. The fea-
Vol. 16 No. 6 December 1998 Japanese Brief Pain Inventory 369 sibility of multinational studies of cancer pain, including patients from Japan and patients where other previously validated language versions of the BPI exist, is supported by the data from this study. Acknowledgments This research was supported by Grants CA26582 from the National Cancer Institute, Public Health Service, and a special Institutional Grant from the American Cancer Society (ACS) SIG #21. References 1. Serlin RC, Mendoza TR, Nakamura Y, Edwards KR, Cleeland CS. When is cancer pain mild moderate or severe? Grading pain severity by its interference with function. Pain 1995;61:277 284. 2. Cleeland CS, Nakamura Y, Mendoza TR, Edwards KR, Douglas J, Serlin RC. Dimensions of the impact of cancer in a four country sample: new information from multidimensional scaling. Pain 1996; 67 (2 3):267 273. 3. Cleeland CS, Gonin R, Hatfield AK, Edmonson JH, Blum RH, Stewart JA, Pandya KJ. Pain and its treatment in outpatients with metastatic cancer. N Engl J Med 1994;330:592 596. 4. Larue F, Colleau SM, Brasseur L, Cleeland CS. Multicentre study of pain and its treatment in France. Br Med J 1995;310(6986):1034 1037. 5. Breitbart W, Rosenfeld BD, Passik SD, Mc- Donald MV, Thaler H, Portenoy RK. The undertreatment of pain in ambulatory AIDS patients. Pain 1996;65:243 249. 6. World Health Organization. Cancer pain relief: with a guide to opioid availability. 2nd ed. Geneva: WHO, 1996. 7. Cleeland CS, Ladinsky JL, Serlin RC, Nugyen CT. Multidimensional measurement of cancer pain: comparisons of US and Vietnamese patients. J Pain Symptom Manage 1988;3:23 27. 8. Wang XS, Mendoza TR, Gao SZ, Cleeland CS. 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. 9. Fuse Y, Fujita N. [The assessment and management of cancer pain] (Japanese). Gan to Kagaku Ryoho (Jpn J Cancer Chemother) 1992;19(3):286 293. 10. Harman HH. Modern factor analysis. 2nd ed. Chicago: University of Chicago Press, 1967. 11. Nunnally JC. Psychometric theory. New York: McGraw-Hill, 1978. 12. Gorsuch RL. Factor analysis. Saunders: Philadelphia, 1974. 13. Takeda F, Uki J. Recent progress in cancer pain management and palliative care in Japan. Ann Acad Med, Singapore 1994;23 (2):296 299. 14. VonRoenn JH, Cleeland CS, Gonin R, Hatfield AK, Pandya, KJ. Physician attitudes and practice in cancer pain management: a survey from the Eastern Cooperative Oncology Group. Ann Intern Med 1993; 119(2):121 126. (Appendixes 1 & 2 follow next page)
370 Uki et al. Vol. 16 No. 6 December 1998 Appendix 1
Vol. 16 No. 6 December 1998 Japanese Brief Pain Inventory 371 Appendix 1 (continued)
372 Uki et al. Vol. 16 No. 6 December 1998 Appendix 2
Vol. 16 No. 6 December 1998 Japanese Brief Pain Inventory 373 Appendix 2 (continued)