This study is the first to determine CPs for mild, moderate, and severe pain in a sample of patients with OA of the hip or knee who are scheduled for joint replacement surgery. The optimal CPs for average pain, as assessed with the BPI and categorized as mild, moderate, and severe were CP(4,6) for patients with OA of the hip and (4,7) for those with OA of the knee, respectively. Hence, the optimal CPs differed between the two patient categories.
Another notable finding was that pain severity ratings were not associated with the duration of pain in either patient group. The optimal CPs for average pain resulted in the highest association with WOMAC pain for both patient categories and among the highest with the bodily pain scale of the SF-36, which supports the validity of the derived optimal CPs.
Previous work in non-cancer pain has reported different CPs for phantom limb pain (4,7), back pain (4,6), and for pain in general (3,6) [7
], and that the optimal CPs for low back pain and OA are (5,8), and (5,7), respectively with the lower CPs of 5 (≤ 5 and > 5) being the most replicable and discriminative [8
]. Hence the cutoffs seem to differ according to disease entities. Several reasons may account for the differences in CPs observed across these studies. For example the etiologies of the pain were different across the various studies, the demographic characteristics of the populations differed, and there were variations in whether ratings of average or worst pain were used to derive the CPs. In the present study, we used average pain intensity without a time dimension to derive CPs, as in previous studies [7
]. Most recently, Paul and Zelman et al. using a homogenous sample of oncology outpatients with pain from bone metastasis, found that when a full range of CPs was tested, using both average and worst pain scores, the optimal CPs were (4,7) [5
]. In addition, only three of the studies tested the full range of possible CPs [5
]. The finding of different CPs for patients with OA of the hip and those with OA of the knee is new and may have several explanations. For example, it may be related to differences in the 'natural history' of the progression of OA of the hip or knee, differences in stage of the disease, differences in etiologies of the pain, or differences in demographic characteristics such as age or sex.
In the present study, we validated the BPI with the derived optimal CPs by assessing associations with the pain scales of the WOMAC and SF-36 questionnaires, using alternative CPs. The association of average pain with WOMAC pain was highest when using the derived optimal CPs, and among the highest with SF-36 bodily pain. In the framing of the items of the questionnaires, the time perspective for the BPI was the past 24 hours, for the WOMAC 48 hours, and for the SF-36 4 weeks. Therefore, a closer association of BPI average pain with WOMAC scores would be expected, and we would put more emphasis on this for the purpose of validation. This difference in time frame between the BPI and the WOMAC may possibly influence the results, however, we think the difference between 24 h and 48 h should be minor. Further, we asked the subjects to respond to questionnaires in relationship to the hip or knee joint that was to be replaced, however if subjects have multiple painful areas, they may have problems assessing the impact of one painful area on function.
In a previous study of patients with unspecified OA, who were undergoing treatment with cyclooxygenase-2 specific inhibitors and non-steroidal anti-inflammatory drugs [33
], average pain exhibited good convergent validity, as assessed by the association with the pain subscale on the WOMAC. Hence, BPI appeared to be a useful measure of pain and pain interference and can be used to assess osteoarthritis pain and functioning.
In non-cancer pain, only the studies by Zelman et al. attempted to validate the severity categories using other outcome measures [5
]. They reported that CPs calculations may be measure-dependent, for example they may be influenced by the differential sensitivity of the functional disability measure for levels of pain severity, or the correlation between pain severity and the reference measure. Further, the choice of pain item will influence the CPs, and it is claimed that pain-related interference may be most appropriate item for determining CPs for pain severity [12
Worsening pain is the main indication for joint replacement surgery [34
], although there are large variations in physicians' judgment regarding the priority of patients for joint replacement surgery. However, there is controversy in the literature about the optimal timing for joint replacement surgery in relationship to the duration of pain and mobility problems [26
]. Therefore, we think the finding of little association between the pain severity ratings and the duration of pain and mobility problems in both the hip and knee patient groups in the present study is notable. Moreover, it is interesting to note that patients with OA of the knee reported longer duration of pain than patients with OA of the hip, which suggests a difference in the time-course of the conditions, or just a difference in selection of patients for surgery.
As noted in some previous reports, pain has been suggested to be a better predictor of disability and the need for surgery than does radiographic evidence of disease [1
]. However, whether a structured pain assessment tool, such as the BPI, is useful in prioritizing patients for joint replacement surgery is not yet clear and needs to be assessed in longitudinal studies.
The patients in the present study were representative of the population of patients with OA who were scheduled for hip or knee joint replacement surgery, and they had similar WOMAC scores as patients in previous studies [2
Several limitations of this study must be noted. Since patients were recruited by mail, the reasons why patients chose not to participate are not known. However, the patients who chose to participate were approximately 3.5 years younger then those who did not participate which may have biased the sample towards the reporting of higher pain intensity scores [36
]. Another limitation was the lack of radiographic categorization of the extent of OA in these patients and the lack of information on comorbidity. In the present study, we did not consider the full range of possible CPs, but concentrated on a range from 3,5 to 5,7, in line with findings in previous reports [5
]. Finally, patients were recruited from the public health care system, which does not necessarily allow for generalization to patients in the private health care system. However, in Norway the public health care system accounts for > 95% of hip and knee replacement surgery, hence we do not think this limits generalization of the results.
Some critique of the concept and use of CPs for pain should be raised. We have shown that the optimal CPs as derived with the standard methods differ between patient categories, which is also shown in previous studies. Therefore, there seems to be no universal optimal CPs, which makes such a classification difficult to use in practice. Because the CPs are based on the optimal group boundaries, this may not represent the best CPs for an individual patient [12
], hence the concept of CPs may be difficult to use in a clinical setting. The categories mild, moderate and severe pain may be easier to interpret than the 11-level 0–10 scale of the BPI. Therefore, the categorization of BPI pain scores as mild, moderate or severe pain according to the optimal CPs may be useful in communication with patients about the interpretation of such scores. However, the severity levels of BPI average pain represent a reduction of the 11-level classification to a 3-level scale, which involves a considerable lack of information and also limits the choice of statistical methods for data analysis. Therefore, for research purposes it is probably better to use BPI on the continuous 0–10 scale. It also seems illogical to include those with a score of 0 (no pain) in a group with mild pain, and in some populations this would lead to large floor effects. Finally, optimal CPs may differ according to age, gender, race, or culture, which could represent another problem with optimal CPs. However, the latter has to our knowledge not been investigated.