- Email: [email protected]
Categorizing the severity of cancer pain: further exploration of the establishment of cutpoints
Pain 113 (2005) 37–44 www.elsevier.com/locate/pain
Categorizing the severity of cancer pain: further exploration of the establishment of cutpoints Steven M. Paula, Diane C. Zelmanb, Meredith Smithc, Christine Miaskowskia,* a
Department of Physiological Nursing, University of California, 2 Koret Way—N631Y, San Francisco, CA 94143-0610, USA b California School of Professional Psychology, Alliant International University, Alameda, CA, USA c Health Economics and Outcomes Research, Purdue Pharma, LP, Stamford, CT, USA Received 16 December 2003; received in revised form 3 September 2004; accepted 13 September 2004
Abstract Previous work by Serlin and colleagues [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;61:277–84] established cutpoints for mild, moderate, and severe cancer pain based on the pain’s level of interference with function. Recent work [Jensen MP, Smith DG, Ehde DM, Robinson LR. Pain site and the effects of amputation pain: further clarification of the meaning of mild, moderate, and severe pain. Pain 2001;91:317–22; Zelman DC, Hoffman DL, Seifeldin R, Dukes, E. Development of a metric for a day of manageable pain control: derivation of pain severity cutpoints for low back pain and osteoarthritis. Pain 2003;106(1/2):35–42] found differences in cutpoints for pain severity for different painrelated conditions. Reasons for these discrepancies may relate to the methods used to determine the cutpoints or to differences based on the type or the cause of the pain. The purposes of this study were to determine the optimal cutpoints for mild, moderate, and severe pain based on patients’ ratings of average and worst pain severity, using a larger range of potential cutpoints, and to determine if those cutpoints distinguished among the three pain severity groups on several outcome measures. Results from a homogenous sample of oncology outpatients with pain from bone metastasis confirm a non-linear relationship between cancer pain severity and interference with function and also confirm that the boundary between a mild and a moderate level of cancer pain is at 4 on a 0–10 numeric rating scale. However, this analysis did not confirm the boundary between moderate and severe cancer pain previously described by Serlin and colleagues [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;61:277–84]. In addition, these results were not consistent with the cutpoints that were found for back pain, phantom limb pain, pain ‘in general’, or osteoarthritis pain reported by Jensen and colleagues and Zelman and colleagues [Jensen MP, Smith DG, Ehde DM, Robinson LR. Pain site and the effects of amputation pain: further clarification of the meaning of mild, moderate, and severe pain. Pain 2001;91:317–22; Zelman DC, Hoffman DL, Seifeldin R, Dukes, E. Development of a metric for a day of manageable pain control: derivation of pain severity cutpoints for low back pain and osteoarthritis. Pain 2003;106(1/2):35–42]. Possible explanations for these differences are discussed, as well as implications for future research. q 2004 Published by Elsevier B.V. on behalf of International Association for the Study of Pain. Keywords: Cutpoints for pain severity; Cancer pain; Bone metastasis
1. Introduction Work by Serlin et al. (1995) established cutpoints for cancer pain severity based on the pain’s level of interference with function (i.e. 0–4 (mild), 5–6 (moderate), and 7–10
* Corresponding author. Tel.: C1 415 476 9407; fax: C1 415 476 8899. E-mail address: [email protected] (C. Miaskowski).
(severe)). However, recent work (Jensen et al., 2001; Zelman et al., 2003) found differences in cutpoints for pain severity for different pain-related conditions. Jensen et al. (2001) found similar cutpoints for back pain (cutpoints were 4,6), but not for phantom limb pain (cutpoints were 4,7), or pain ‘in general’ (cutpoints were 3,6). Zelman et al. (2003) found cutpoints of 5,8 for low back pain and 5,7 for osteoarthritis pain, with the lower cutpoint of 5 (i.e. %5 and O5) being the most replicable and discriminative cutpoint. However, in
0304-3959/$20.00 q 2004 Published by Elsevier B.V. on behalf of International Association for the Study of Pain. doi:10.1016/j.pain.2004.09.014
38
S.M. Paul et al. / Pain 113 (2005) 37–44
contrast to Serlin et al. (1995) who used ratings of worst pain to establish the cutpoints, both Jensen et al. (2001) and Zelman et al. (2003) used ratings of average pain in their analyses. Several questions, that have implications for the determination of cutpoints remain unanswered in the studies published to date. In both the Jensen et al. (2001) and Serlin et al. (1995) studies, cutpoints were created based on a categorization of whole number responses. For example, Serlin et al.’s classification (1995) determined that a single score of 6 would be categorized as ‘moderate’ pain and a score of 7 as ‘severe’ pain. However, patients may report pain intensity scores that are not whole numbers or researchers may want to categorize a mean pain intensity rating from a week’s worth of ratings. Therefore, it is important to establish absolute cutpoints so that a score of 6.5 can be classified appropriately. Another issue is the absolute number of cutpoint solutions that were considered in the investigation. For example, Zelman et al. (2003) tested a full range of cutpoints to determine the optimal set of cutpoints that discriminated between different levels of pain severity, while Jensen et al. (2001) and Serlin et al. (1995) used a limited range of cutpoint options. Yet another issue is that the naming of the cutpoints as mild, moderate, and severe was done based on relating the rating of pain intensity to a single outcome and verification of these cutpoints with other outcomes would add to our understanding of the validity of the cutpoints. Finally, one of the questions that remains unanswered is whether a single set of cutpoints applies to all types (e.g. nociceptive, neuropathic) or causes of pain. The findings from previous studies (Jensen et al., 2001; Zelman et al., 2003) suggest that different cutpoints may need to be determined for different types or causes of acute and chronic pain. Therefore, the current study sought to address several of these unanswered questions, using a homogenous sample of oncology outpatients with pain from bone metastasis. Specifically, the purposes of this study were to determine the optimal cutpoints for mild, moderate, and severe pain based on patients’ ratings of average and worst pain, using a larger range of potential cutpoints, and to determine if those cutpoints distinguished among the three pain severity groups on several outcome measures (i.e. functional status, mood states and QOL).
experiencing pain from bone metastasis, were recruited from seven outpatient settings in Northern California including: a universitybased cancer center, two community-based oncology practices, one health maintenance organization, one outpatient radiation therapy center, one veteran’s administration facility, and one military hospital. One hundred and seventy-four patients completed the study and of those patients, 160 who had complete data on baseline ratings of pain intensity and pain interference from the Brief Pain Inventory (BPI; Cleeland, 1991) were included in this analysis. The participants were adult oncology outpatients (O18 years) who were able to read, write, and understand English; had a Karnofsky Performance Status (KPS) score (Karnofsky, 1977) of R50; had an average pain intensity score of R2.5; and had radiographic evidence of bone metastasis. 2.2. Instruments At baseline, patients completed a demographic questionnaire, the KPS rating, measures of pain intensity, the interference items from the BPI, the Profile of Mood States (POMS), and the Medical Outcomes Survey-Short Form (MOS-SF36), and their medical records were reviewed for cancer diagnosis and treatment information. 2.2.1. Numeric rating scales (NRSs) of pain intensity Patients were asked to rate their present pain intensity, as well as average, worst, and least pain using 0 (no pain) to 10 (excruciating) NRSs that are known to be valid and reliable measures of pain intensity (Downie et al., 1978; Huskisson, 1974; Jensen, 2003). The coefficient alpha for the pain intensity scale composed of the four pain intensity items was 0.81. 2.2.2. Interference items from the BPI The interference scale of the BPI (Cleeland, 1991) measures how pain interferes with seven domains using an 11-point scale that ranges from 0 (does not interfere) to 10 (completely interferes). The seven domains are general activity, mood, walking ability, sleep, enjoyment of life, normal work, and relations with others. An eighth domain sexual activity was added in this research study. To maintain consistency with previous studies that determined cutpoints for pain severity (Jensen et al., 2001; Serlin et al., 1995; Zelman et al., 2003), the sexual activity item was not incorporated into the cutpoint analysis. However, the differences among the three pain severity groups in patients’ ratings of pain’s level of interference with sexual activity are reported in Section 3.
2.1. Sample and settings
2.2.3. Profile of mood states The shortened version of the Profile of Mood States, developed by Shacham (1983) measures six independent mood or affective states (i.e. anxiety, depression, anger, vigor, fatigue, confusion). In addition, a total mood disturbance score is calculated by summing the scores of the six mood subscales with the vigor subscale weighted negatively.
This study is a secondary analysis of data from a large, randomized clinical trial (RCT) that tested the effectiveness of a self-care intervention compared to standard care in improving the management of cancer pain (Miaskowski et al., 2004; West et al., 2003). Two hundred and twelve oncology outpatients, who were
2.2.4. Multidimensional quality of life scale-cancer (MQOLS-CA2) A shortened version of the MQOLS-CA2 (Dibble et al., 1998) was used in this study. This 17-item instrument asked patients to rate various aspects of their QOL using 0–10 NRSs. A total QOL score was calculated with higher scores indicating a better QOL.
2. Methods
S.M. Paul et al. / Pain 113 (2005) 37–44
2.2.5. MOS-SF36 The MOS-SF consists of 36 items and is a well-validated measure of health status (Ware and Sherbourne, 1992). The instrument yields eight separate subscales and two composite measures of physical and mental functioning. 2.3. Data collection procedures This study was approved by the Committee on Human Research at the University of California, San Francisco and at each of the study sites. The details of this intervention study are described elsewhere (Miaskowski et al., 2004; West et al., 2003). In brief, after providing written, informed consent, patients were randomized into the intervention or standard care groups. Patients in both groups were seen in their homes and asked to complete the baseline measures used in this analysis.
39
items as indicated by Pillai’s trace, Wilk’s lambda, and Hotelling’s trace F statistics. In order to determine if the optimal cutpoints for pain severity distinguished among the three pain severity levels (i.e. mild, moderate and severe) on clinical outcome measures (i.e. functional status, mood states and QOL), a one-way ANOVA was done with post hoc comparisons performed using the Scheffe procedure. All calculations used actual values. Adjustments were not made for missing data. Therefore, the cohort for each analysis was dependent on the largest complete set of data across groups. For all tests, a P-value of !0.05 was considered statistically significant.
3. Results 3.1. Sample demographics
2.4. Data analysis Descriptive statistics and frequency distributions were generated for the patients’ demographic and disease-related characteristics. Cutpoints that divided the sample of patients into mild, moderate, or severe pain were created for ratings of average and worst pain using the analytic strategy described by Serlin et al. (1995). Eight different categorical variables, that represented the eight possible combinations for the cutpoints, between 3 and 7, were created and related to the set of seven interference items from the BPI using multivariate analysis of variance (MANOVA). For example, cutpoints (CP) 3,5 were coded so that a pain severity rating of 1–3 would correspond to ‘mild’, O3–5 to ‘moderate’, and O5–10 to ‘severe’. The criterion used to determine the optimal set of cutpoints for mild, moderate, and severe pain was that a MANOVA among pain severity categories yielded the largest F ratio for the between-category effect on the seven interference
The demographic and disease characteristics of the patients are summarized in Tables 1 and 2, respectively. The majority of the patients were female (71.3%), with a mean age of 59.44 years, an average of 15.04 years of education, and an average KPS score of 70.47. 3.2. Cutpoint calculations As shown in Table 3, for average pain, CP 4,7 (i.e. 1–4 is mild pain, O4–7 is moderate pain, and O7–10 is severe pain) were the optimal cutpoints, in that they had the largest between-category-F-ratios, using Pillai’s trace, Wilk’s lambda, and Hotelling’s statistic. Using average pain intensity scores, 51.9% of the sample (nZ83) was classified as having mild pain, 45.6% (nZ73) as having moderate pain, and 2.5% (nZ4) as having severe pain.
Table 1 Differences in demographic characteristics among the three groups of patients based on the optimal cutpoint for worst pain intensity Characteristic
Total sample (NZ160), mean (SD)
Mild pain 1–4 (NZ19), mean (SD)
Moderate pain O4–7 (NZ68), mean (SD)
Severe pain O7–10 (NZ73), mean (SD)
Age (years) Education (years) KPS score*
59.44 (12.40) 15.04 (3.16) 70.47 (11.58)
62.89 (14.53) 14.42 (3.56) 74.47 (12.57)
60.29 (12.39) 15.15 (3.37) 71.99 (11.69)
57.72 (11.69) 15.11 (2.86) 67.99 (10.08)
%
%
%
%
28.8 71.3 28.3
31.6 68.4 15.8
27.9 72.1 25.0
28.8 71.3 34.7
55.6 44.4
68.4 31.6
50.0 50.0
57.5 42.5
88.7 11.3
78.9 21.1
89.7 10.3
90.3 9.7
40.3 27.7 17.6 14.4
57.8 15.8 26.4 0.0
41.2 22.1 17.6 19.1
34.7 36.1 15.3 13.9
Gender Male Female Lives alone Marital status Married/partnered Other Ethnicity Caucasian Other Employment status Retired Disability Full or part-time Other
KPS, Karnofsky performance status; SD, standard deviation. *PZ0.03.
40
S.M. Paul et al. / Pain 113 (2005) 37–44
Table 2 Differences in disease and treatment characteristics among the three groups of patients based on the optimal cutpoint for worst pain intensity Characteristic Diagnosis Breast Prostate Lung Other Current cancer treatmenta Chemotherapy Hormonal therapy Radiation therapy Biotherapy No treatment
Total sample (NZ160) (%)
Mild pain 1–4 (NZ19) (%)
Moderate pain O4–7 (NZ68) (%)
Severe pain O7–10 (NZ73) (%)
51.9 10.8 10.8 26.5
52.6 10.5 0.0 36.9
50.7 9.0 14.9 28.4
52.8 12.5 9.7 25.0
46.2 32.3 17.7 2.5 12.3
31.6 36.8 26.3 0.0 10.5
52.2 34.3 17.9 0.0 12.3
41.7 29.2 15.3 5.6 12.9
a Since patients could be receiving more than one type of current treatment, separate Chi Square analyses were done, for each of the treatments, to test for differences among groups.
For worst pain, CP 4,7 (i.e. 1–4 is mild pain, O4–7 is moderate pain, and O7–10 is severe pain) were also the optimal cutpoints, in that they had the largest betweencategory-F-ratios, using Pillai’s trace, and Wilk’s lambda. Using worst pain intensity scores, 11.9% of the sample (nZ19) was classified as having mild pain, 42.5% (nZ68) as having moderate pain, and 45.6% (nZ73) as having severe pain. In addition, a significant positive correlation was found between worst pain score and total interference score on the BPI (rZ0.54, P!0.0001). Consequently, the CP 4,7 was determined to be the optimal choice and was used for subsequent analyses. 3.3. Differences in demographic, disease, and treatment characteristics As shown in Tables 1 and 2, no differences were found among the three pain severity groups in any demographic, disease, or treatment characteristics except KPS score. Post hoc contrasts revealed that the mean KPS scores for the patients with mild and moderate pain, collapsed together, were significantly higher than the mean KPS score for the patients with severe pain.
Table 3 Results of the multivariate analysis of variance to determine optimal cutpoints using average and worst pain intensity scores and the interference items from the Brief Pain Inventory Cutpoints
Pillai’s trace
Wilk’s lambda
Hotelling’s trace
Rank
F
Rank
F
Rank
F
2.687 3.129 3.327 2.933 3.281 3.622 1.451 1.573
6 4 3 5 2 1 8 7
2.792 3.208 3.371 3.076 3.403 3.705 1.448 1.566
6 4 3 5 2 1 8 7
2.895 3.287 3.414 3.218 3.525 3.787 1.444 1.558
4.662 4.323 4.391 4.566 4.457 4.778 3.831 4.314
2 5 6 3 4 1 8 7
4.979 4.644 4.569 4.966 4.862 5.037 4.092 4.539
2 5 7 1 4 3 8 6
5.296 4.965 4.747 5.367 5.269 5.294 4.353 4.763
Average pain CPA3,5 6 CPA3,6 4 CPA3,7 2 CPA4,5 5 CPA4,6 3 CPA4,7 1 CPA5,6 8 CPA5,7 7 Worst pain CP3,5 2 CP3,6 6 CP3,7 5 CP4,5 3 CP4,6 4 CP4,7 1 CP5,6 8 CP5,7 7
3.4. Differences in pain intensity and pain interference scores As illustrated in Fig. 1, for pain now, average pain, and worst pain, significant differences in pain intensity scores were found among the three pain intensity groups (all P!0.001, respectively). Post hoc contrasts demonstrated that patients in the severe pain group had significantly higher pain intensity scores than either of the other two pain groups and that patients in the moderate pain group had significantly higher pain intensity scores than patients in the mild pain group. For least pain, significant differences in scores were found among the three pain intensity groups (PZ0.001). Post hoc contrasts
Fig. 1. Differences in pain now, average pain, worst pain, and least pain among the three pain severity groups. All values are plotted as meansGSD.
S.M. Paul et al. / Pain 113 (2005) 37–44
41
Table 4 Mean, standard deviation, and 95% confidence intervals for the overall interference score for each severity category based on the optimal cutpoints for worst pain Severity category
Mean
Standard deviation
95% Confidence intervals
Mild pain Moderate pain Severe pain
1.89 4.50 5.66
2.06 1.94 1.88
0.90–2.89 4.03–4.97 5.22–6.10
demonstrated a significant difference in scores only between the mild and the severe pain groups. Table 4 lists the mean, standard deviation, and 95% confidence intervals for the total pain interference score for each pain intensity group. Fig. 2 illustrates differences in the interference scores for each of the individual items on the BPI, as well as for the mean interference score. Statistically significant between group differences were found for all of the individual interference items on the BPI, as well as for the total interference scores (all P!0.001). Post hoc contrasts revealed significant differences among the three groups (i.e. mild!moderate!severe) for activity, mood, sleep, enjoyment of life, and total interference score (Fig. 2A). Post hoc contrasts revealed significant differences between the mild, but not between the moderate and severe pain groups (i.e. mild!moderate, mild!severe, moderateZsevere) for walking ability and normal work interference scores (Fig. 2B). Finally, post hoc contrasts revealed significant differences for severe pain, but not for mild compared to moderate pain (i.e. mildZmoderate, mild !severe, moderate !severe) for relations with other people and sexual activity interference scores (Fig. 2C). 3.5. Differences in mood states, QOL, and functional status scores As shown in Table 5, significant differences were found among the three pain intensity groups in anger, fatigue, and TMD scores from the POMS. In addition, QOL scores were significantly different among the three pain intensity groups. As shown in Table 6, significant differences were found among the three pain intensity groups for the following scales on the MOS-SF36: role-physical, bodily pain, vitality, social functioning, role-emotional, mental health index, standardized physical component score, and standardized mental health component score.
4. Discussion This study is the first to attempt to confirm the cutpoints for mild, moderate, and severe pain in a sample of oncology outpatients with pain from bone metastasis. Using the methods described by Serlin et al. (1995), our results confirm a non-linear relationship between cancer pain
Fig. 2. Differences in interference scores for each of the individual items on the Brief Pain Inventory, as well as for the total interference score. All values are plotted as meansGSD.
severity and interference with function. In addition, this analysis provides additional confirmation that the boundary between a mild and a moderate level of cancer pain is at 4. However, this analysis did not confirm the boundary between moderate and severe cancer pain. In the Serlin et al. article (1995), the boundary between moderate and severe pain was between the values of 6 and 7 with ratings of 7 being in the severe category. In this analysis, the boundary between moderate and severe pain was at 7, with ratings of 7 being in the moderate category and ratings of O7 being in the severe category.
42
S.M. Paul et al. / Pain 113 (2005) 37–44
Table 5 Differences in profile of mood state scores and quality of life scores among the three groups of patients based on the optimal cutpoint for worst pain intensity Subscale and TMD scores
Total sample (NZ160), mean (SD)
Mild pain 1–4 (NZ19), mean (SD)
Moderate pain O4–7 (NZ68), mean (SD)
Severe pain O7–10 (NZ73), mean (SD)
F statistic, P-value
Anxiety Depression Angera Vigor Fatigueb Confusion TMD Scorea Total MQOLS scoreb
5.65 (4.34) 4.48 (4.07) 3.65 (3.58) 5.47 (3.83) 8.94 (4.85) 5.54 (4.85) 22.79 (18.06) 65.17 (15.76)
3.89 (3.74) 2.95 (3.42) 2.05 (2.74) 7.16 (2.85) 5.53 (4.39) 5.26 (2.62) 12.53 (15.40) 74.48 (13.36)
5.31 (3.86) 4.47 (3.74) 3.35 (3.16) 5.24 (3.76) 8.67 (4.02) 5.34 (3.21) 21.91 (14.61) 65.04 (14.07)
6.43 (4.76) 4.89 (4.45) 4.35 (3.99) 5.25 (4.05) 10.08 (5.24) 5.78 (3.74) 26.28 (20.53) 62.88 (17.07)
FZ3.03, PZ0.051 FZ1.74, PZ0.18 FZ3.62, PZ0.029 FZ2.11, PZ0.12 FZ7.35, PZ0.001 FZ0.36, PZ0.70 FZ4.72, PZ0.01 FZ4.26, PZ0.016
TMD, total mood disturbance; MQOLS, multidimensional quality of life scale. a Mild!severe, mildZmoderate, moderateZsevere. b Mild!moderate, mild!severe, moderateZsevere.
One possible explanation for the discrepancy between the findings from this study and the study by Serlin et al. (1995) is that they combined data from four different countries (i.e. United States, France, China and Philippines) and included an additional criterion that the MANOVA results yielded fairly consistent results across the four nations. In fact, Serlin et al. (1995) attempted to minimize the differences across the nations, because the ultimate criterion that was used to establish the cutpoints was that the most useful cutpoints would produce the largest ratio of between-category (i.e. the three pain levels) and interaction (i.e. pain by nation) F statistics. However, if one examines only the pain level F statistics, CP 4,7 would be optimal. In the Serlin et al. study, it is only by using the ratio criterion that the CP 4,6 cutpoint appears optimal. Since the sample for this study was primarily Caucasian, one cannot determine from these data whether cutpoints vary based on cultural or ethnic differences in how individuals interpret pain intensity ratings. This point warrants investigation in future studies.
Another possible explanation for the differences in cutpoints between the two studies is that patients in the study by Serlin et al. (1995) were recruited from both inpatient and outpatient settings and the patients’ cancer pain had multiple etiologies. The present analysis was based on a sample of oncology outpatients with a homogeneous pain problem (i.e. pain from bone metastasis). While one could suggest that the findings from the study by Serlin et al. (1995) are more generalizable to the population of patients with cancer pain, than the findings from this study, additional research is warranted with patients with different types of cancer pain to determine the precise delineations of cutpoints for pain severity. One potential explanation for differences in the cutpoints for pain severity, based on the etiology of the pain, is that different types or causes of pain may have a differential impact on various aspects of function. Certainly, bone metastasis, particularly to the weight bearing joints, can result in marked impairments in patients’ ability to work and to walk. Therefore, if interference with function is
Table 6 Differences in MOS-SF36 scores among the three groups of patients based on the optimal cutpoint for worst pain intensity MOS-SF36 scores
Total sample (NZ160), mean (SD)
Mild pain 1–4 (NZ19), mean (SD)
Moderate pain O4–7 (NZ68), mean (SD)
Severe pain O7–10 (NZ73), mean (SD)
F statistic, P-value
Physical functioning Role-physicala Bodily painb General health perception Vitalitya Social functioninga Role-emotionalc Mental health indexa Standardized physical componenta Standardized mental componentd
34.19 (23.14) 9.84 (23.31) 35.86 (17.23) 41.50 (20.79)
41.17 (24.45) 25.00 (35.36) 56.74 (21.30) 51.16 (21.79)
34.58 (25.63) 9.19 (21.98) 37.31 (14.20) 39.07 (19.31)
32.00 (20.08) 6.51 (19.11) 29.07 (13.82) 41.25 (21.36)
FZ1.20, PZ0.303 FZ5.04, PZ0.008 FZ26.13, P!0.001 FZ2.57, PZ0.079
35.97 (18.95) 52.58 (27.37) 39.58 (43.79) 66.24 (17.47) 30.07 (6.77)
47.11 (21.17) 69.74 (22.94) 56.14 (44.52) 77.90 (16.13) 34.51 (7.10)
35.47 (18.30) 53.49 (26.61) 44.11 (45.14) 67.18 (17.31) 29.59 (7.18)
33.54 (18.18) 47.26 (27.50) 31.05 (40.95) 62.33 (16.68) 29.36 (5.91)
FZ4.06, PZ0.019 FZ5.44, PZ0.005 FZ3.20, PZ0.044 FZ6.59, PZ0.002 FZ4.87, PZ0.009
44.45 (11.43)
50.87 (10.51)
45.31 (11.57)
41.98 (10.87)
FZ5.15, PZ0.007
a b c d
Mild!moderate, mild!severe, moderateZsevere. Mild!moderate!severe. Unable to discern which differences are statistically significant. Mild!severe, mildZmoderate, moderateZsevere.
S.M. Paul et al. / Pain 113 (2005) 37–44
the outcome used to determine pain severity cutpoints, it is plausible that the painful condition, with its associated comorbidities, could influence various aspects of function differentially and drive severity cutpoints up or down. Likewise, if a different outcome (e.g. mood or QOL) was used to establish or verify pain severity cutpoints, the impact that the specific pain problem or associated comorbidities has on the outcome measure, independent of pain intensity, will effect the establishment of the cutpoint categories. Our cutpoint findings were similar to those reported by Jensen et al. (2001) for phantom limb pain, but not for back pain (i.e. 4,6) or pain ‘in general’ (i.e. 3,6). Nor were they similar to those reported by Zelman et al. (2003) for low back pain (i.e. 5,8) or osteoarthritis (i.e. 5,7). However, findings from this study cannot be compared directly to those of Jensen et al. (2001) and Zelman et al. (2003), because in addition to the use of different patient populations, both studies used ratings of average pain intensity to determine the cutpoints for mild, moderate, and severe pain. Of note, in this study of oncology outpatients with pain from bone metastasis, identical cutpoints for pain severity were found using average and worst pain intensity scores (i.e. mild painZ1–4, moderate painZO4–7, severe painZO7–10). However, when average pain was used to establish the cutpoints CP 4,7, only 2.5% of the sample was categorized with severe pain in contrast to 45.6% of the sample when worst pain was used in the cutpoints analysis. Whether to use ratings of average or worst pain for the establishment of cutpoints warrants further investigation and may need to be evaluated in terms of the type of pain and the common comorbidities associated with the specific pain problem. Zelman and colleagues’ (2003) rationale for using average pain in their cutpoints analysis was that because their sample consisted of individuals with chronic noncancer-related pain (i.e. low back pain and osteoarthritis) that their daily experience of pain was comparatively more stable than cancer patients who experience persistent as well as breakthrough pain; and that ratings of average pain intensity better reflect the experiences that most likely lead to disruptions in function and mood. While the choice of which pain intensity rating is used in the derivation of cutpoints may appear to be a subtle distinction, it has important implications for research and clinical practice because cutpoints are used to select pain treatments, to develop pain treatment algorithms in clinical practice guidelines, or to determine the effectiveness of pain management interventions (Cleeland et al., 1998; Hwang et al., 2002; National Comprehensive Cancer Network and American Cancer Society, 2001). The cutpoints derived in this study were able to distinguish among the three severity groups on some, but not on all of the outcome measures that were evaluated (see Tables 4–6). It should be noted that significant differences were found in the mean total interference scores among the three severity groups and that the 95% confidence intervals did not overlap. This finding suggests that the cutpoints for
43
the three pain severity groups are reasonably sensitive and specific. However, when the scores on the individual interference items were compared among the three severity groups (see Fig. 2), only activity, mood, sleep, and enjoyment of life distinguished among the three groups (i.e. mild!moderate!severe). These findings suggest that different patient outcomes may be influenced differentially by different levels of pain severity or may be dependent on the measure used or influenced by the sensitivity of the measure. Likewise, the differences, and equally important the lack of differences, among the severity groups on various outcome measures may represent true clinical differences that are dependent on the influence of a specific pain problem on a particular outcome measure. As more research is done on comparisons among severity groups on different clinical outcomes and on the establishment of cutpoints using a variety of patient outcomes, these differences in outcomes may become appropriate measures to use to judge the effectiveness of various pain management interventions. The findings from the four studies that have explored the establishment of cutpoints for mild, moderate, and severe pain using average and worst pain intensity scores (Jensen et al., 2001; Serlin et al., 1995; Zelman et al., 2003) suggest several areas for future investigation. The choice of whether to use ratings of average or worst pain to establish cutpoints for pain severity may need to be based on the distribution of the severity categories within the sample; the specific components of the pain problem being investigated (e.g. persistent stable pain versus persistent pain with breakthrough pain) or the etiology of the pain problem (e.g. nociceptive pain, neuropathic pain or pain of mixed etiology). In addition, verification of the severity cutpoints derived from ratings of average and worst pain, using additional outcome measures (e.g. mood states, QOL) may provide additional information for clinical practice and research. Since severity categories are used widely in clinical practice to assist in treatment decisions, additional research is warranted to determine if the use of ratings of average or worst pain to establish cutpoints contributes to significant differences in patient outcomes or in approaches to pain treatment.
Acknowledgements This study was supported by a grant (CA 64734) from the National Cancer Institute and by Purdue Pharma, L.P.
References Cleeland CS. Pain assessment in cancer. In: Osoba D, editor. Effect of cancer on quality of life. Boca Raton, FL: CRC Press; 1991. p. 293–305. Cleeland CS, Nakamura Y, Mendoza TR, Edwards KR, Douglas J, Serlin RC. Dimensions of the impact of cancer pain in a four country
44
S.M. Paul et al. / Pain 113 (2005) 37–44
sample: new information from multidimensional scaling. Pain 1998; 67(2/3):267–73. Dibble SL, Padilla G, Dodd MJ, Miaskowski C. Gender differences in the dimensions of quality of life. Oncol Nurs Forum 1998;25(3):577–83. Downie WW, Leatham PA, Rhind VM, Wright V, Branco JA, Anderson JA. Studies with pain rating scales. Ann Rheum Dis 1978;37(4):378–81. Huskisson EC. Measurement of pain. Lancet 1974;2:1127–31. Hwang SS, Chang VT, Kasimis B. Dynamic cancer pain management outcomes: the relationship between pain severity, pain relief, functional interference, satisfaction, and global quality of life over time. J Pain Symptom Manage 2002;23(3):190–200. Jensen MP. The validity and reliability of pain measures in adults with cancer. J Pain 2003;4(1):2–21. Jensen MP, Smith DG, Ehde DM, Robinson LR. Pain site and the effects of amputation pain: further clarification of the meaning of mild, moderate, and severe pain. Pain 2001;91:317–22. Karnofsky D. Performance scale. In: Kennealey GT, Mitchell MS, editors. Factors that influence the therapeutic response to cancer. New York: Plenum Press; 1977. p. 206–22. Miaskowski C, Dodd MJ, West C, Schumacher K, Paul SM, Tripathy D, Koo P. A randomized clinical trial of the effectiveness of the
PRO-SELFq Pain Control Program in improving cancer pain management. J Clin Oncol 2004;22(9):1712–30. National Comprehensive Cancer Network and American Cancer Society. Cancer pain-treatment guidelines for patients. Version I. Atlanta, GA: American Cancer Society; 2001. 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;61:277–84. Shacham S. A shortened version of the profile of mood states. J Pers Assess 1983;47:305–6. Ware Jr JE, Sherbourne CD. MOS 36-item short-form health survey (SF36). I. Conceptual framework and item selection. Med Care 1992;30: 473–83. West CM, Dodd MJ, Paul SM, Schumacher K, Tripathy D, Koo P, Miaskowski C. PRO-SELFq Pain Control Program: an effective approach for cancer pain management. Oncol Nurs Forum 2003; 30(1):65–73. Zelman DC, Hoffman DL, Seifeldin R, Dukes E. Development of a metric for a day of manageable pain control: derivation of pain severity cut-points for low back pain and osteoarthritis. Pain 2003;106(1/2): 35–42.
Categorizing the severity of cancer pain: further exploration of the establishment of cutpoints Steven M. Paula, Diane C. Zelmanb, Meredith Smithc, Christine Miaskowskia,* a
Department of Physiological Nursing, University of California, 2 Koret Way—N631Y, San Francisco, CA 94143-0610, USA b California School of Professional Psychology, Alliant International University, Alameda, CA, USA c Health Economics and Outcomes Research, Purdue Pharma, LP, Stamford, CT, USA Received 16 December 2003; received in revised form 3 September 2004; accepted 13 September 2004
Abstract Previous work by Serlin and colleagues [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;61:277–84] established cutpoints for mild, moderate, and severe cancer pain based on the pain’s level of interference with function. Recent work [Jensen MP, Smith DG, Ehde DM, Robinson LR. Pain site and the effects of amputation pain: further clarification of the meaning of mild, moderate, and severe pain. Pain 2001;91:317–22; Zelman DC, Hoffman DL, Seifeldin R, Dukes, E. Development of a metric for a day of manageable pain control: derivation of pain severity cutpoints for low back pain and osteoarthritis. Pain 2003;106(1/2):35–42] found differences in cutpoints for pain severity for different painrelated conditions. Reasons for these discrepancies may relate to the methods used to determine the cutpoints or to differences based on the type or the cause of the pain. The purposes of this study were to determine the optimal cutpoints for mild, moderate, and severe pain based on patients’ ratings of average and worst pain severity, using a larger range of potential cutpoints, and to determine if those cutpoints distinguished among the three pain severity groups on several outcome measures. Results from a homogenous sample of oncology outpatients with pain from bone metastasis confirm a non-linear relationship between cancer pain severity and interference with function and also confirm that the boundary between a mild and a moderate level of cancer pain is at 4 on a 0–10 numeric rating scale. However, this analysis did not confirm the boundary between moderate and severe cancer pain previously described by Serlin and colleagues [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;61:277–84]. In addition, these results were not consistent with the cutpoints that were found for back pain, phantom limb pain, pain ‘in general’, or osteoarthritis pain reported by Jensen and colleagues and Zelman and colleagues [Jensen MP, Smith DG, Ehde DM, Robinson LR. Pain site and the effects of amputation pain: further clarification of the meaning of mild, moderate, and severe pain. Pain 2001;91:317–22; Zelman DC, Hoffman DL, Seifeldin R, Dukes, E. Development of a metric for a day of manageable pain control: derivation of pain severity cutpoints for low back pain and osteoarthritis. Pain 2003;106(1/2):35–42]. Possible explanations for these differences are discussed, as well as implications for future research. q 2004 Published by Elsevier B.V. on behalf of International Association for the Study of Pain. Keywords: Cutpoints for pain severity; Cancer pain; Bone metastasis
1. Introduction Work by Serlin et al. (1995) established cutpoints for cancer pain severity based on the pain’s level of interference with function (i.e. 0–4 (mild), 5–6 (moderate), and 7–10
* Corresponding author. Tel.: C1 415 476 9407; fax: C1 415 476 8899. E-mail address: [email protected] (C. Miaskowski).
(severe)). However, recent work (Jensen et al., 2001; Zelman et al., 2003) found differences in cutpoints for pain severity for different pain-related conditions. Jensen et al. (2001) found similar cutpoints for back pain (cutpoints were 4,6), but not for phantom limb pain (cutpoints were 4,7), or pain ‘in general’ (cutpoints were 3,6). Zelman et al. (2003) found cutpoints of 5,8 for low back pain and 5,7 for osteoarthritis pain, with the lower cutpoint of 5 (i.e. %5 and O5) being the most replicable and discriminative cutpoint. However, in
0304-3959/$20.00 q 2004 Published by Elsevier B.V. on behalf of International Association for the Study of Pain. doi:10.1016/j.pain.2004.09.014
38
S.M. Paul et al. / Pain 113 (2005) 37–44
contrast to Serlin et al. (1995) who used ratings of worst pain to establish the cutpoints, both Jensen et al. (2001) and Zelman et al. (2003) used ratings of average pain in their analyses. Several questions, that have implications for the determination of cutpoints remain unanswered in the studies published to date. In both the Jensen et al. (2001) and Serlin et al. (1995) studies, cutpoints were created based on a categorization of whole number responses. For example, Serlin et al.’s classification (1995) determined that a single score of 6 would be categorized as ‘moderate’ pain and a score of 7 as ‘severe’ pain. However, patients may report pain intensity scores that are not whole numbers or researchers may want to categorize a mean pain intensity rating from a week’s worth of ratings. Therefore, it is important to establish absolute cutpoints so that a score of 6.5 can be classified appropriately. Another issue is the absolute number of cutpoint solutions that were considered in the investigation. For example, Zelman et al. (2003) tested a full range of cutpoints to determine the optimal set of cutpoints that discriminated between different levels of pain severity, while Jensen et al. (2001) and Serlin et al. (1995) used a limited range of cutpoint options. Yet another issue is that the naming of the cutpoints as mild, moderate, and severe was done based on relating the rating of pain intensity to a single outcome and verification of these cutpoints with other outcomes would add to our understanding of the validity of the cutpoints. Finally, one of the questions that remains unanswered is whether a single set of cutpoints applies to all types (e.g. nociceptive, neuropathic) or causes of pain. The findings from previous studies (Jensen et al., 2001; Zelman et al., 2003) suggest that different cutpoints may need to be determined for different types or causes of acute and chronic pain. Therefore, the current study sought to address several of these unanswered questions, using a homogenous sample of oncology outpatients with pain from bone metastasis. Specifically, the purposes of this study were to determine the optimal cutpoints for mild, moderate, and severe pain based on patients’ ratings of average and worst pain, using a larger range of potential cutpoints, and to determine if those cutpoints distinguished among the three pain severity groups on several outcome measures (i.e. functional status, mood states and QOL).
experiencing pain from bone metastasis, were recruited from seven outpatient settings in Northern California including: a universitybased cancer center, two community-based oncology practices, one health maintenance organization, one outpatient radiation therapy center, one veteran’s administration facility, and one military hospital. One hundred and seventy-four patients completed the study and of those patients, 160 who had complete data on baseline ratings of pain intensity and pain interference from the Brief Pain Inventory (BPI; Cleeland, 1991) were included in this analysis. The participants were adult oncology outpatients (O18 years) who were able to read, write, and understand English; had a Karnofsky Performance Status (KPS) score (Karnofsky, 1977) of R50; had an average pain intensity score of R2.5; and had radiographic evidence of bone metastasis. 2.2. Instruments At baseline, patients completed a demographic questionnaire, the KPS rating, measures of pain intensity, the interference items from the BPI, the Profile of Mood States (POMS), and the Medical Outcomes Survey-Short Form (MOS-SF36), and their medical records were reviewed for cancer diagnosis and treatment information. 2.2.1. Numeric rating scales (NRSs) of pain intensity Patients were asked to rate their present pain intensity, as well as average, worst, and least pain using 0 (no pain) to 10 (excruciating) NRSs that are known to be valid and reliable measures of pain intensity (Downie et al., 1978; Huskisson, 1974; Jensen, 2003). The coefficient alpha for the pain intensity scale composed of the four pain intensity items was 0.81. 2.2.2. Interference items from the BPI The interference scale of the BPI (Cleeland, 1991) measures how pain interferes with seven domains using an 11-point scale that ranges from 0 (does not interfere) to 10 (completely interferes). The seven domains are general activity, mood, walking ability, sleep, enjoyment of life, normal work, and relations with others. An eighth domain sexual activity was added in this research study. To maintain consistency with previous studies that determined cutpoints for pain severity (Jensen et al., 2001; Serlin et al., 1995; Zelman et al., 2003), the sexual activity item was not incorporated into the cutpoint analysis. However, the differences among the three pain severity groups in patients’ ratings of pain’s level of interference with sexual activity are reported in Section 3.
2.1. Sample and settings
2.2.3. Profile of mood states The shortened version of the Profile of Mood States, developed by Shacham (1983) measures six independent mood or affective states (i.e. anxiety, depression, anger, vigor, fatigue, confusion). In addition, a total mood disturbance score is calculated by summing the scores of the six mood subscales with the vigor subscale weighted negatively.
This study is a secondary analysis of data from a large, randomized clinical trial (RCT) that tested the effectiveness of a self-care intervention compared to standard care in improving the management of cancer pain (Miaskowski et al., 2004; West et al., 2003). Two hundred and twelve oncology outpatients, who were
2.2.4. Multidimensional quality of life scale-cancer (MQOLS-CA2) A shortened version of the MQOLS-CA2 (Dibble et al., 1998) was used in this study. This 17-item instrument asked patients to rate various aspects of their QOL using 0–10 NRSs. A total QOL score was calculated with higher scores indicating a better QOL.
2. Methods
S.M. Paul et al. / Pain 113 (2005) 37–44
2.2.5. MOS-SF36 The MOS-SF consists of 36 items and is a well-validated measure of health status (Ware and Sherbourne, 1992). The instrument yields eight separate subscales and two composite measures of physical and mental functioning. 2.3. Data collection procedures This study was approved by the Committee on Human Research at the University of California, San Francisco and at each of the study sites. The details of this intervention study are described elsewhere (Miaskowski et al., 2004; West et al., 2003). In brief, after providing written, informed consent, patients were randomized into the intervention or standard care groups. Patients in both groups were seen in their homes and asked to complete the baseline measures used in this analysis.
39
items as indicated by Pillai’s trace, Wilk’s lambda, and Hotelling’s trace F statistics. In order to determine if the optimal cutpoints for pain severity distinguished among the three pain severity levels (i.e. mild, moderate and severe) on clinical outcome measures (i.e. functional status, mood states and QOL), a one-way ANOVA was done with post hoc comparisons performed using the Scheffe procedure. All calculations used actual values. Adjustments were not made for missing data. Therefore, the cohort for each analysis was dependent on the largest complete set of data across groups. For all tests, a P-value of !0.05 was considered statistically significant.
3. Results 3.1. Sample demographics
2.4. Data analysis Descriptive statistics and frequency distributions were generated for the patients’ demographic and disease-related characteristics. Cutpoints that divided the sample of patients into mild, moderate, or severe pain were created for ratings of average and worst pain using the analytic strategy described by Serlin et al. (1995). Eight different categorical variables, that represented the eight possible combinations for the cutpoints, between 3 and 7, were created and related to the set of seven interference items from the BPI using multivariate analysis of variance (MANOVA). For example, cutpoints (CP) 3,5 were coded so that a pain severity rating of 1–3 would correspond to ‘mild’, O3–5 to ‘moderate’, and O5–10 to ‘severe’. The criterion used to determine the optimal set of cutpoints for mild, moderate, and severe pain was that a MANOVA among pain severity categories yielded the largest F ratio for the between-category effect on the seven interference
The demographic and disease characteristics of the patients are summarized in Tables 1 and 2, respectively. The majority of the patients were female (71.3%), with a mean age of 59.44 years, an average of 15.04 years of education, and an average KPS score of 70.47. 3.2. Cutpoint calculations As shown in Table 3, for average pain, CP 4,7 (i.e. 1–4 is mild pain, O4–7 is moderate pain, and O7–10 is severe pain) were the optimal cutpoints, in that they had the largest between-category-F-ratios, using Pillai’s trace, Wilk’s lambda, and Hotelling’s statistic. Using average pain intensity scores, 51.9% of the sample (nZ83) was classified as having mild pain, 45.6% (nZ73) as having moderate pain, and 2.5% (nZ4) as having severe pain.
Table 1 Differences in demographic characteristics among the three groups of patients based on the optimal cutpoint for worst pain intensity Characteristic
Total sample (NZ160), mean (SD)
Mild pain 1–4 (NZ19), mean (SD)
Moderate pain O4–7 (NZ68), mean (SD)
Severe pain O7–10 (NZ73), mean (SD)
Age (years) Education (years) KPS score*
59.44 (12.40) 15.04 (3.16) 70.47 (11.58)
62.89 (14.53) 14.42 (3.56) 74.47 (12.57)
60.29 (12.39) 15.15 (3.37) 71.99 (11.69)
57.72 (11.69) 15.11 (2.86) 67.99 (10.08)
%
%
%
%
28.8 71.3 28.3
31.6 68.4 15.8
27.9 72.1 25.0
28.8 71.3 34.7
55.6 44.4
68.4 31.6
50.0 50.0
57.5 42.5
88.7 11.3
78.9 21.1
89.7 10.3
90.3 9.7
40.3 27.7 17.6 14.4
57.8 15.8 26.4 0.0
41.2 22.1 17.6 19.1
34.7 36.1 15.3 13.9
Gender Male Female Lives alone Marital status Married/partnered Other Ethnicity Caucasian Other Employment status Retired Disability Full or part-time Other
KPS, Karnofsky performance status; SD, standard deviation. *PZ0.03.
40
S.M. Paul et al. / Pain 113 (2005) 37–44
Table 2 Differences in disease and treatment characteristics among the three groups of patients based on the optimal cutpoint for worst pain intensity Characteristic Diagnosis Breast Prostate Lung Other Current cancer treatmenta Chemotherapy Hormonal therapy Radiation therapy Biotherapy No treatment
Total sample (NZ160) (%)
Mild pain 1–4 (NZ19) (%)
Moderate pain O4–7 (NZ68) (%)
Severe pain O7–10 (NZ73) (%)
51.9 10.8 10.8 26.5
52.6 10.5 0.0 36.9
50.7 9.0 14.9 28.4
52.8 12.5 9.7 25.0
46.2 32.3 17.7 2.5 12.3
31.6 36.8 26.3 0.0 10.5
52.2 34.3 17.9 0.0 12.3
41.7 29.2 15.3 5.6 12.9
a Since patients could be receiving more than one type of current treatment, separate Chi Square analyses were done, for each of the treatments, to test for differences among groups.
For worst pain, CP 4,7 (i.e. 1–4 is mild pain, O4–7 is moderate pain, and O7–10 is severe pain) were also the optimal cutpoints, in that they had the largest betweencategory-F-ratios, using Pillai’s trace, and Wilk’s lambda. Using worst pain intensity scores, 11.9% of the sample (nZ19) was classified as having mild pain, 42.5% (nZ68) as having moderate pain, and 45.6% (nZ73) as having severe pain. In addition, a significant positive correlation was found between worst pain score and total interference score on the BPI (rZ0.54, P!0.0001). Consequently, the CP 4,7 was determined to be the optimal choice and was used for subsequent analyses. 3.3. Differences in demographic, disease, and treatment characteristics As shown in Tables 1 and 2, no differences were found among the three pain severity groups in any demographic, disease, or treatment characteristics except KPS score. Post hoc contrasts revealed that the mean KPS scores for the patients with mild and moderate pain, collapsed together, were significantly higher than the mean KPS score for the patients with severe pain.
Table 3 Results of the multivariate analysis of variance to determine optimal cutpoints using average and worst pain intensity scores and the interference items from the Brief Pain Inventory Cutpoints
Pillai’s trace
Wilk’s lambda
Hotelling’s trace
Rank
F
Rank
F
Rank
F
2.687 3.129 3.327 2.933 3.281 3.622 1.451 1.573
6 4 3 5 2 1 8 7
2.792 3.208 3.371 3.076 3.403 3.705 1.448 1.566
6 4 3 5 2 1 8 7
2.895 3.287 3.414 3.218 3.525 3.787 1.444 1.558
4.662 4.323 4.391 4.566 4.457 4.778 3.831 4.314
2 5 6 3 4 1 8 7
4.979 4.644 4.569 4.966 4.862 5.037 4.092 4.539
2 5 7 1 4 3 8 6
5.296 4.965 4.747 5.367 5.269 5.294 4.353 4.763
Average pain CPA3,5 6 CPA3,6 4 CPA3,7 2 CPA4,5 5 CPA4,6 3 CPA4,7 1 CPA5,6 8 CPA5,7 7 Worst pain CP3,5 2 CP3,6 6 CP3,7 5 CP4,5 3 CP4,6 4 CP4,7 1 CP5,6 8 CP5,7 7
3.4. Differences in pain intensity and pain interference scores As illustrated in Fig. 1, for pain now, average pain, and worst pain, significant differences in pain intensity scores were found among the three pain intensity groups (all P!0.001, respectively). Post hoc contrasts demonstrated that patients in the severe pain group had significantly higher pain intensity scores than either of the other two pain groups and that patients in the moderate pain group had significantly higher pain intensity scores than patients in the mild pain group. For least pain, significant differences in scores were found among the three pain intensity groups (PZ0.001). Post hoc contrasts
Fig. 1. Differences in pain now, average pain, worst pain, and least pain among the three pain severity groups. All values are plotted as meansGSD.
S.M. Paul et al. / Pain 113 (2005) 37–44
41
Table 4 Mean, standard deviation, and 95% confidence intervals for the overall interference score for each severity category based on the optimal cutpoints for worst pain Severity category
Mean
Standard deviation
95% Confidence intervals
Mild pain Moderate pain Severe pain
1.89 4.50 5.66
2.06 1.94 1.88
0.90–2.89 4.03–4.97 5.22–6.10
demonstrated a significant difference in scores only between the mild and the severe pain groups. Table 4 lists the mean, standard deviation, and 95% confidence intervals for the total pain interference score for each pain intensity group. Fig. 2 illustrates differences in the interference scores for each of the individual items on the BPI, as well as for the mean interference score. Statistically significant between group differences were found for all of the individual interference items on the BPI, as well as for the total interference scores (all P!0.001). Post hoc contrasts revealed significant differences among the three groups (i.e. mild!moderate!severe) for activity, mood, sleep, enjoyment of life, and total interference score (Fig. 2A). Post hoc contrasts revealed significant differences between the mild, but not between the moderate and severe pain groups (i.e. mild!moderate, mild!severe, moderateZsevere) for walking ability and normal work interference scores (Fig. 2B). Finally, post hoc contrasts revealed significant differences for severe pain, but not for mild compared to moderate pain (i.e. mildZmoderate, mild !severe, moderate !severe) for relations with other people and sexual activity interference scores (Fig. 2C). 3.5. Differences in mood states, QOL, and functional status scores As shown in Table 5, significant differences were found among the three pain intensity groups in anger, fatigue, and TMD scores from the POMS. In addition, QOL scores were significantly different among the three pain intensity groups. As shown in Table 6, significant differences were found among the three pain intensity groups for the following scales on the MOS-SF36: role-physical, bodily pain, vitality, social functioning, role-emotional, mental health index, standardized physical component score, and standardized mental health component score.
4. Discussion This study is the first to attempt to confirm the cutpoints for mild, moderate, and severe pain in a sample of oncology outpatients with pain from bone metastasis. Using the methods described by Serlin et al. (1995), our results confirm a non-linear relationship between cancer pain
Fig. 2. Differences in interference scores for each of the individual items on the Brief Pain Inventory, as well as for the total interference score. All values are plotted as meansGSD.
severity and interference with function. In addition, this analysis provides additional confirmation that the boundary between a mild and a moderate level of cancer pain is at 4. However, this analysis did not confirm the boundary between moderate and severe cancer pain. In the Serlin et al. article (1995), the boundary between moderate and severe pain was between the values of 6 and 7 with ratings of 7 being in the severe category. In this analysis, the boundary between moderate and severe pain was at 7, with ratings of 7 being in the moderate category and ratings of O7 being in the severe category.
42
S.M. Paul et al. / Pain 113 (2005) 37–44
Table 5 Differences in profile of mood state scores and quality of life scores among the three groups of patients based on the optimal cutpoint for worst pain intensity Subscale and TMD scores
Total sample (NZ160), mean (SD)
Mild pain 1–4 (NZ19), mean (SD)
Moderate pain O4–7 (NZ68), mean (SD)
Severe pain O7–10 (NZ73), mean (SD)
F statistic, P-value
Anxiety Depression Angera Vigor Fatigueb Confusion TMD Scorea Total MQOLS scoreb
5.65 (4.34) 4.48 (4.07) 3.65 (3.58) 5.47 (3.83) 8.94 (4.85) 5.54 (4.85) 22.79 (18.06) 65.17 (15.76)
3.89 (3.74) 2.95 (3.42) 2.05 (2.74) 7.16 (2.85) 5.53 (4.39) 5.26 (2.62) 12.53 (15.40) 74.48 (13.36)
5.31 (3.86) 4.47 (3.74) 3.35 (3.16) 5.24 (3.76) 8.67 (4.02) 5.34 (3.21) 21.91 (14.61) 65.04 (14.07)
6.43 (4.76) 4.89 (4.45) 4.35 (3.99) 5.25 (4.05) 10.08 (5.24) 5.78 (3.74) 26.28 (20.53) 62.88 (17.07)
FZ3.03, PZ0.051 FZ1.74, PZ0.18 FZ3.62, PZ0.029 FZ2.11, PZ0.12 FZ7.35, PZ0.001 FZ0.36, PZ0.70 FZ4.72, PZ0.01 FZ4.26, PZ0.016
TMD, total mood disturbance; MQOLS, multidimensional quality of life scale. a Mild!severe, mildZmoderate, moderateZsevere. b Mild!moderate, mild!severe, moderateZsevere.
One possible explanation for the discrepancy between the findings from this study and the study by Serlin et al. (1995) is that they combined data from four different countries (i.e. United States, France, China and Philippines) and included an additional criterion that the MANOVA results yielded fairly consistent results across the four nations. In fact, Serlin et al. (1995) attempted to minimize the differences across the nations, because the ultimate criterion that was used to establish the cutpoints was that the most useful cutpoints would produce the largest ratio of between-category (i.e. the three pain levels) and interaction (i.e. pain by nation) F statistics. However, if one examines only the pain level F statistics, CP 4,7 would be optimal. In the Serlin et al. study, it is only by using the ratio criterion that the CP 4,6 cutpoint appears optimal. Since the sample for this study was primarily Caucasian, one cannot determine from these data whether cutpoints vary based on cultural or ethnic differences in how individuals interpret pain intensity ratings. This point warrants investigation in future studies.
Another possible explanation for the differences in cutpoints between the two studies is that patients in the study by Serlin et al. (1995) were recruited from both inpatient and outpatient settings and the patients’ cancer pain had multiple etiologies. The present analysis was based on a sample of oncology outpatients with a homogeneous pain problem (i.e. pain from bone metastasis). While one could suggest that the findings from the study by Serlin et al. (1995) are more generalizable to the population of patients with cancer pain, than the findings from this study, additional research is warranted with patients with different types of cancer pain to determine the precise delineations of cutpoints for pain severity. One potential explanation for differences in the cutpoints for pain severity, based on the etiology of the pain, is that different types or causes of pain may have a differential impact on various aspects of function. Certainly, bone metastasis, particularly to the weight bearing joints, can result in marked impairments in patients’ ability to work and to walk. Therefore, if interference with function is
Table 6 Differences in MOS-SF36 scores among the three groups of patients based on the optimal cutpoint for worst pain intensity MOS-SF36 scores
Total sample (NZ160), mean (SD)
Mild pain 1–4 (NZ19), mean (SD)
Moderate pain O4–7 (NZ68), mean (SD)
Severe pain O7–10 (NZ73), mean (SD)
F statistic, P-value
Physical functioning Role-physicala Bodily painb General health perception Vitalitya Social functioninga Role-emotionalc Mental health indexa Standardized physical componenta Standardized mental componentd
34.19 (23.14) 9.84 (23.31) 35.86 (17.23) 41.50 (20.79)
41.17 (24.45) 25.00 (35.36) 56.74 (21.30) 51.16 (21.79)
34.58 (25.63) 9.19 (21.98) 37.31 (14.20) 39.07 (19.31)
32.00 (20.08) 6.51 (19.11) 29.07 (13.82) 41.25 (21.36)
FZ1.20, PZ0.303 FZ5.04, PZ0.008 FZ26.13, P!0.001 FZ2.57, PZ0.079
35.97 (18.95) 52.58 (27.37) 39.58 (43.79) 66.24 (17.47) 30.07 (6.77)
47.11 (21.17) 69.74 (22.94) 56.14 (44.52) 77.90 (16.13) 34.51 (7.10)
35.47 (18.30) 53.49 (26.61) 44.11 (45.14) 67.18 (17.31) 29.59 (7.18)
33.54 (18.18) 47.26 (27.50) 31.05 (40.95) 62.33 (16.68) 29.36 (5.91)
FZ4.06, PZ0.019 FZ5.44, PZ0.005 FZ3.20, PZ0.044 FZ6.59, PZ0.002 FZ4.87, PZ0.009
44.45 (11.43)
50.87 (10.51)
45.31 (11.57)
41.98 (10.87)
FZ5.15, PZ0.007
a b c d
Mild!moderate, mild!severe, moderateZsevere. Mild!moderate!severe. Unable to discern which differences are statistically significant. Mild!severe, mildZmoderate, moderateZsevere.
S.M. Paul et al. / Pain 113 (2005) 37–44
the outcome used to determine pain severity cutpoints, it is plausible that the painful condition, with its associated comorbidities, could influence various aspects of function differentially and drive severity cutpoints up or down. Likewise, if a different outcome (e.g. mood or QOL) was used to establish or verify pain severity cutpoints, the impact that the specific pain problem or associated comorbidities has on the outcome measure, independent of pain intensity, will effect the establishment of the cutpoint categories. Our cutpoint findings were similar to those reported by Jensen et al. (2001) for phantom limb pain, but not for back pain (i.e. 4,6) or pain ‘in general’ (i.e. 3,6). Nor were they similar to those reported by Zelman et al. (2003) for low back pain (i.e. 5,8) or osteoarthritis (i.e. 5,7). However, findings from this study cannot be compared directly to those of Jensen et al. (2001) and Zelman et al. (2003), because in addition to the use of different patient populations, both studies used ratings of average pain intensity to determine the cutpoints for mild, moderate, and severe pain. Of note, in this study of oncology outpatients with pain from bone metastasis, identical cutpoints for pain severity were found using average and worst pain intensity scores (i.e. mild painZ1–4, moderate painZO4–7, severe painZO7–10). However, when average pain was used to establish the cutpoints CP 4,7, only 2.5% of the sample was categorized with severe pain in contrast to 45.6% of the sample when worst pain was used in the cutpoints analysis. Whether to use ratings of average or worst pain for the establishment of cutpoints warrants further investigation and may need to be evaluated in terms of the type of pain and the common comorbidities associated with the specific pain problem. Zelman and colleagues’ (2003) rationale for using average pain in their cutpoints analysis was that because their sample consisted of individuals with chronic noncancer-related pain (i.e. low back pain and osteoarthritis) that their daily experience of pain was comparatively more stable than cancer patients who experience persistent as well as breakthrough pain; and that ratings of average pain intensity better reflect the experiences that most likely lead to disruptions in function and mood. While the choice of which pain intensity rating is used in the derivation of cutpoints may appear to be a subtle distinction, it has important implications for research and clinical practice because cutpoints are used to select pain treatments, to develop pain treatment algorithms in clinical practice guidelines, or to determine the effectiveness of pain management interventions (Cleeland et al., 1998; Hwang et al., 2002; National Comprehensive Cancer Network and American Cancer Society, 2001). The cutpoints derived in this study were able to distinguish among the three severity groups on some, but not on all of the outcome measures that were evaluated (see Tables 4–6). It should be noted that significant differences were found in the mean total interference scores among the three severity groups and that the 95% confidence intervals did not overlap. This finding suggests that the cutpoints for
43
the three pain severity groups are reasonably sensitive and specific. However, when the scores on the individual interference items were compared among the three severity groups (see Fig. 2), only activity, mood, sleep, and enjoyment of life distinguished among the three groups (i.e. mild!moderate!severe). These findings suggest that different patient outcomes may be influenced differentially by different levels of pain severity or may be dependent on the measure used or influenced by the sensitivity of the measure. Likewise, the differences, and equally important the lack of differences, among the severity groups on various outcome measures may represent true clinical differences that are dependent on the influence of a specific pain problem on a particular outcome measure. As more research is done on comparisons among severity groups on different clinical outcomes and on the establishment of cutpoints using a variety of patient outcomes, these differences in outcomes may become appropriate measures to use to judge the effectiveness of various pain management interventions. The findings from the four studies that have explored the establishment of cutpoints for mild, moderate, and severe pain using average and worst pain intensity scores (Jensen et al., 2001; Serlin et al., 1995; Zelman et al., 2003) suggest several areas for future investigation. The choice of whether to use ratings of average or worst pain to establish cutpoints for pain severity may need to be based on the distribution of the severity categories within the sample; the specific components of the pain problem being investigated (e.g. persistent stable pain versus persistent pain with breakthrough pain) or the etiology of the pain problem (e.g. nociceptive pain, neuropathic pain or pain of mixed etiology). In addition, verification of the severity cutpoints derived from ratings of average and worst pain, using additional outcome measures (e.g. mood states, QOL) may provide additional information for clinical practice and research. Since severity categories are used widely in clinical practice to assist in treatment decisions, additional research is warranted to determine if the use of ratings of average or worst pain to establish cutpoints contributes to significant differences in patient outcomes or in approaches to pain treatment.
Acknowledgements This study was supported by a grant (CA 64734) from the National Cancer Institute and by Purdue Pharma, L.P.
References Cleeland CS. Pain assessment in cancer. In: Osoba D, editor. Effect of cancer on quality of life. Boca Raton, FL: CRC Press; 1991. p. 293–305. Cleeland CS, Nakamura Y, Mendoza TR, Edwards KR, Douglas J, Serlin RC. Dimensions of the impact of cancer pain in a four country
44
S.M. Paul et al. / Pain 113 (2005) 37–44
sample: new information from multidimensional scaling. Pain 1998; 67(2/3):267–73. Dibble SL, Padilla G, Dodd MJ, Miaskowski C. Gender differences in the dimensions of quality of life. Oncol Nurs Forum 1998;25(3):577–83. Downie WW, Leatham PA, Rhind VM, Wright V, Branco JA, Anderson JA. Studies with pain rating scales. Ann Rheum Dis 1978;37(4):378–81. Huskisson EC. Measurement of pain. Lancet 1974;2:1127–31. Hwang SS, Chang VT, Kasimis B. Dynamic cancer pain management outcomes: the relationship between pain severity, pain relief, functional interference, satisfaction, and global quality of life over time. J Pain Symptom Manage 2002;23(3):190–200. Jensen MP. The validity and reliability of pain measures in adults with cancer. J Pain 2003;4(1):2–21. Jensen MP, Smith DG, Ehde DM, Robinson LR. Pain site and the effects of amputation pain: further clarification of the meaning of mild, moderate, and severe pain. Pain 2001;91:317–22. Karnofsky D. Performance scale. In: Kennealey GT, Mitchell MS, editors. Factors that influence the therapeutic response to cancer. New York: Plenum Press; 1977. p. 206–22. Miaskowski C, Dodd MJ, West C, Schumacher K, Paul SM, Tripathy D, Koo P. A randomized clinical trial of the effectiveness of the
PRO-SELFq Pain Control Program in improving cancer pain management. J Clin Oncol 2004;22(9):1712–30. National Comprehensive Cancer Network and American Cancer Society. Cancer pain-treatment guidelines for patients. Version I. Atlanta, GA: American Cancer Society; 2001. 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;61:277–84. Shacham S. A shortened version of the profile of mood states. J Pers Assess 1983;47:305–6. Ware Jr JE, Sherbourne CD. MOS 36-item short-form health survey (SF36). I. Conceptual framework and item selection. Med Care 1992;30: 473–83. West CM, Dodd MJ, Paul SM, Schumacher K, Tripathy D, Koo P, Miaskowski C. PRO-SELFq Pain Control Program: an effective approach for cancer pain management. Oncol Nurs Forum 2003; 30(1):65–73. Zelman DC, Hoffman DL, Seifeldin R, Dukes E. Development of a metric for a day of manageable pain control: derivation of pain severity cut-points for low back pain and osteoarthritis. Pain 2003;106(1/2): 35–42.