Validation of brief pain inventory to Brazilian patients with pain

  • Published on

  • View

  • Download



    Validation of brief pain inventory to Brazilianpatients with pain

    Karine A. Ferreira & Manoel Jacobsen Teixeira &Tito R. Mendonza & Charles S. Cleeland

    Received: 16 October 2009 /Accepted: 16 February 2010 /Published online: 10 March 2010# Springer-Verlag 2010

    AbstractPurpose To validate the Brazilian version of the Brief PainInventory (BPI-B) scale and to determine the optimalcutpoints for mild, moderate, and severe pain based onpatients' rating of their worst pain.Methods One hundred forty-three outpatients with cancerwere recruited in Hospital das ClinicasUniversity of SaoPaulo, Brazil.Results Confirmatory factor analysis confirmed two underly-ing dimensions, pain severity, and pain interference, withCronbach's of 0.91 and 0.87, respectively. Convergentvalidity was shown by the correlation observed between theBPI dimensions with the EORTC-QLQ-C30 pain scale andthe McGill Pain Questionnaire. The BPI-B detected signif-icant differences in the two dimensions by disease andperformance status, supporting known-group validity. For

    the worst pain, the optimal cutpoints were 4 and 7 (14=mild pain, 57=moderate, and 810=severe).Conclusions Our data show that BPI-B is a brief, useful, andvalid tool for assessing pain and its impact on patient's life.

    Keywords Neoplasm . Pain . Pain measurement .

    Reproducibility of results . Validation studies


    Cancer is one of the most prevalent diseases in the world.Most cancer patients had or will have pain during thecourse of their disease [1]. Thus, the regular assessment ofpain is important to early identification and treatment ofcancer.

    Pain is one of the most frequent symptoms in patientswith cancer. In Brazil, almost 30% of cancer patients arebeing treated at any one time, and 6090% of those withadvanced disease have experienced pain [2]. In Braziliancancer centers, from 34.8% to 52.8% of patients havecancer pain, and of them, 14.433.3% have referred severepain and 11.613.2% moderate pain [3]. Almost 50% ofpatients with cancer pain have received inadequate paintreatment during the course of their treatment [4].

    The reasons for such inadequate treatment of pain arenot clear, but some studies have suggested the discrepancybetween patient and physician in judging the severity oftheir patients pain and physician underestimation of theirpatient's pain were also predictive of inadequate painmanagement [1, 5, 6]. In addition, patients with uncon-trolled pain were not likely to have a formal assessment orreassessment of their pain or a pain-care order in thenursing notes, or to have been seen by the hospitalspecialist palliative care team [7]. Perhaps, if the healthcare

    K. A. Ferreira (*)Instituto do Cancer do Estado de Sao PauloICESP,University of Guarulhos-UnG,R. Coronel Camiso, 409, apto. 72, Butanta,Sao Paulo, SP, Brazil05590-120e-mail:

    K. A. Ferreira :M. J. TeixeiraMultidisciplinary Pain Center, Hospital das Clinicas,School of Medicine, University of So Paulo,So Paulo, Brazil

    M. J. TeixeiraDepartment of Neurosurgery, School of Medicine,University of So Paulo,So Paulo, Brazil

    T. R. Mendonza : C. S. CleelandDepartment of Symptom Research,UT-M.D. Anderson Cancer Center,Houston, USA

    Support Care Cancer (2011) 19:505511DOI 10.1007/s00520-010-0844-7

  • providers used a standard self-assessment pain scale, thepatients' pain could be more easily identified and treated.

    If cancer-related pain is to be improved, data on theprevalence, severity, and impact on quality of life of painmust be obtained and the effectiveness of various treat-ments documented. A number of instruments have beendeveloped to measure clinical pain, mainly unidimensionalpain measurement tools [8]. However, few pain measure-ment instruments are available in Brazil. Thus, it is veryimportant to make available an instrument that healthcareproviders can use daily to assess the pain of outpatient andinpatients.

    Two of the most commonly used multidimensional painscales are the Brief Pain Inventory (BPI) and the McGillPain Questionnaire (MPQ). The MPQ has already beentranslated and validated into Brazilian Portuguese [9, 10].While the MPQ mainly measures the qualities of pain, itdoes not assess how pain interferes with patients function-ing. In addition to measuring the intensity and location ofpain and the effectiveness of pain therapy, the BPI alsoevaluates how pain interferes with patients functioning.

    The BPI is one of the most used instruments around theworld. It was recommended to be included in all chronic-pain clinical trials [11]. The BPI was first developed inEnglish [12] and has been used widely in the United States.This scale has been translated into many different languagessuch as Russian [13], Chinese [14], Italian [15], German[16], Greek [17], Spanish [18], and Japanese [5]. The BPI isa short scale and is easy to understand. However, this scalehas not yet been translated and validated into BrazilianPortuguese. Thus, the purpose of this study was to presentthe Brazilian version of the BPI, to evaluate its psychomet-ric properties among Brazilian patients with cancer pain,and to determine the optimal cutpoints (CP) for mild,moderate, and severe pain based on patients' rating of theirworst pain.

    Patients and methods

    We performed a cross-sectional study at the Multidisciplin-ary Pain Center of the Hospital das Clinicas, University ofSao Paulo in Sao Paulo, Brazil, after receiving approval bythe institutions ethics committee. We recruited 143 cancerpatients with pain. They were recruited between June 2004and December 2006.

    Eligible patients were required to (1) be an outpatient atthe institution, (2) have a pathologic diagnosis of cancer,(3) be diagnosed by a pain specialist as having only chroniccancer-related pain and no other painful condition, (4) havehad no surgery within the past 30 days, (5) have notreceived radiation therapy or chemotherapy within the past30 days, (6) do not have any infection, and (7) be 18 years

    of age or older. If an eligible patient agreed to participate,we obtained written informed consent.

    The mean age of the 143 outpatients participating in thestudy was 57.30 (standard deviation [SD] = 13.28, range=2785), and 61.54% (n=88) of them were women. About53.85% (n=77) had completed at least elementary school.Only 16.78% (n=24) were illiterate, and 23.08% (n=33)had completed high school.

    All patients had a pain syndrome resulting from theircancer. The mean time after pain onset was 8.61 months(SD=11.64, median=5.00, range=196). About 41% (n=58) of the patients had a diagnosis of breast cancer, 23.78%(n=34) had lung cancer, and 12% had prostate (n=16) orgenitourinary cancer (n=17). The mean time since cancerdiagnosis was 28.79 months (SD=33.13, median=17.50,range=2168). The majority of the patients had a metasta-sis (65.73%, n=94).


    Subjects were required to complete the Brazilian version ofthe Brief Pain Inventory (BPI-B), the Brazilian version ofthe McGill Pain Questionnaire, and the European Organi-zation for Research and Treatment of Cancer Quality ofLife-core 30 (EORTC QLQ-c30) questionnaire [19]. Inaddition, the demographic and clinical characteristics andthe performance status were also assessed.

    We retrieved data on the patients' diagnosis, diseasestage, and treatment from the hospital medical records.Demographic data were obtained from the patients byinterview. The Karnofsky performance status scale (KPS)was used to assess performance status [20].

    Linguistic adaptation

    The BPI was translated into Brazilian Portuguese using astandard translation and back-translation procedure. TheEnglish-language items were initially translated into Portu-guese by two native Brazilians who spoke fluent English. Acommittee of bilingually fluent Brazilian pain expertsevaluated and approved these item translations. The itemswere then translated back into English by a bilingualtranslator who had not seen the original English version.The English back-translation and the original versions werecompared. Finally, a Brazilian version of the BPI wasproduced.

    The resulting questionnaire was pilot tested amongpatients with cancer and non-cancer pain. On the basis ofthe results from the pilot test, we made further translationrefinements and finalized the tool. The BPI consists of nineitems that are arranged in two dimensions: intensity/severity of pain (sensory dimension, items 3 to 6) andinterference of pain (impact) in the patient's life (reactive

    506 Support Care Cancer (2011) 19:505511

  • dimension, items 9a to 9g) [12]. The BPI asks patients torate their pain intensity and the pain interference (withgeneral activities, mood, walking ability, normal work,relationships with others, sleep, and enjoyment of life) in an11-point scale ranging from 0 (no pain/no interference) to10 (as bad as it can be). In addition, it also includes acorporal diagram to assess pain location (item 2), measuresthe percentage of pain relief (item 8), and asks patients todescribe which treatments are being used to control pain.The scores for the two dimensions range from 0 to 10 andare calculated using the mean of the total items. A highscore represents a high pain intensity or pain interference[1, 12].

    The McGill Pain Questionnaire (MPQ) was translatedinto Brazilian Portuguese by Pimenta and Teixeira in 1996[10]. The long-form McGill Pain Questionnaire includes 78pain descriptors that are arranged in 20 subclasses of wordsthat measure the sensorial-discriminative, affective-motivational, and evaluative-cognitive dimensions of pain[21]. There is a domain called miscellaneous that includesitems that could not be included in the other three domains.Each word from these subclasses has a rank valueindicative of the relative intensity of pain.

    In the current study, the long form of the MPQ wasscored according to the rank values of the wordsselected using the Pain Rating Index for the Sensory,Affective, Evaluative, and Miscellaneous groups, as wellas for the Total and number of words chosen (NWC)[21]. In the current study, The BV-LF-MPQ demonstratedinternal consistency when measured by the Cronbach'salpha coefficient for all 20 subgroups of words (Cron-bach's =0.87) and for sensorial and affective dimensions(Sensorial = 0.73 and Affective = 0.8, respectively). Thereliability was not good for the miscellaneous dimension(Cronbach's =0.54).

    The pain scale of the EORTC QLQ-C30 version 3.0 wasused to assess pain [19]. The QLQ-C30 consists of 30 itemsthat assess health-related quality of life. The pain scaleincluded two items that assess pain intensity (item 9) andpain interference in daily activities (item 19). The scorerange from 0 to 100. A higher score represented a higher(better) level of pain and a higher (worse) level. In oursample, the EORTC QLQ-C30 showed internal consistencyin functional scales (Cronbach's =0.74) and symptomscales (Cronbach's =0.72).

    Statistical analysis

    The internal consistency reliability of the BPI-B wasevaluated by Cronbach's coefficient alpha. This coefficientwas computed for the interference and severity scales of theBPI. A value of 0.70 or greater was considered adequate forgroup comparisons [22].

    Construct validity was determined by confirmatoryfactor analysis (CFA). CFA was conducted using structur-al equation modeling in LISREL version 8.8 for Windows[23]. The hypothesized model was based on the underly-ing structure of the English version of the BPI [1]. Indicesused to evaluate model fit in CFA included the goodness-of-fit index (GFI), normed fit index (NFI), adjustedgoodness-of-fit index (AGFI), non-normed fit index(NNFI), comparative fit index (CFI), normed chi-square(chi-square/degree of freedom), and root-mean-squareerror of approximation (RMSEA). If the models are fit,their GFI, NFI, NNFI, CFI, GFI, and AGFI should behigher than 0.90, the RMSEA lower than 0.10, and thenormed Chi-square lower than five [24].

    The Spearman's correlation coefficients were calculatedto assess convergent validity between the BPI-B, MPQ, andQLQ-c30 pain scale.

    Known-group validity was evaluated by comparingsubgroups of patients known to differ on clinical variablesusing the Student's t or MannWhitney U tests depends ontheir adherence to normal distribution. We hypothesizedthat patients with a higher performance status and withlocal or locoregional disease would report better function-ing and less pain than patients with a lower performancestatus and with metastatic disease.

    We also evaluated the possible boundaries, or cutpoints,of the BPI-B for defining mild, moderate, and severe pain,based on patient's ratings of their worst pain. On the basisof other studies that examined different cutpoints [2528],we tested seven cutpoint models. We created sevendifferent categorical variables that represented seven possi-ble combinations for the cutpoints between 3 and 7 andrelated them to the set of seven interference items (9a to 9g)from the BPI-B using multivariate analysis of variance(MANOVA). The criterion used to determine the bettercutpoints to discriminate between levels of pain severitywas that a MANOVA among pain severity categoriesyielded the largest F ratio for the category effect on theseven interference items as indicated by Pillai's trace, Wilk'slambda, and Hotelling's trace F values.

    For all tests, a P value of

  • closely followed by general activity, enjoyment of life,and sleep.

    Construct validity

    A confirmatory factor analysis (CFA) was used to deter-mine the adjustment of the BPI-B factors identified in theEnglish version: pain interference (reactive dimension) and

    pain severity/intensity (sensory dimension). The hypothe-sized model is presented in Fig. 1. The independent modelthat tests the hypothesis that all variables are uncorrelatedwas rejected in BPI-B model, normed 2=1.83, P=0.00.The BPI-B model with two dimensions (pain severity andpain impact/interference) showed a good fit to data: GFI=0.82, CFI=0.95, NFI=0.91, NNFI=0.94 and RMSEA=0.11 (CI95%=0.0720.15).

    Fig. 1 Confirmatory factoranalysis for the Brief PainInventory-B

    Variables Mean Median SD Cronbach's if item deleted

    Severity items (pain) 0.91a

    Least 4.13 4.00 3.09 0.87

    Average 2.43 2.00 2.52 0.85

    Now 4.79 5.00 2.54 0.87

    Worst 7.08 7.00 2.62 0.92

    Interference items 0.87a

    General activity 6.87 8.00 3.25 0.83

    Mood 4.96 6.00 3.81 0.87

    Walking ability 5.08 5.50 3.54 0.84

    Normal work 7.03 8.00 3.34 0.84

    Relationships with others 3.32 2.00 3.63 0.86

    Sleep 5.15 6.00 3.88 0.86

    Enjoyment of life 5.98 8.00 3.93 0.84

    Table 1 Descriptive analyses ofthe BPI items and reliabilityanalysis

    a Cronbach's alpha of the component

    508 Support Care Cancer (2011) 19:505511

  • Reliability

    The reliability of the BPI-B was evaluated by the internalconsistency (Cronbach's Coefficient). We separatelycalculated coefficient alpha for the pain severity items andthe pain interference items. Cronbach's alpha coefficientresults are presented in Table 1. The internal consistency forthe pain severity dimension was 0.91 and for the paininterference...


View more >