Abstract
Objective
To examine the association of concurrent low back pain (LBP), and other musculoskeletal pain comorbidity, with knee pain severity in symptomatic knee osteoarthritis (OA).
Methods
1389 individuals from the Progression Cohort of the Osteoarthritis Initiative, age 45-79 with symptomatic tibiofemoral knee OA, were studied. Participants identified pain in the low back, neck, shoulder, elbow, wrist, hand, hip, knee, ankle, or foot. The primary outcome was the pain subscale of the Western Ontario McMaster Universities Osteoarthritis Index (WOMAC) applied to the more symptomatic knee. We examined WOMAC pain score in persons with and without LBP, before and after adjusting for other musculoskeletal symptoms.
Results
57.4% of participants reported LBP. WOMAC pain score (possible range 0-20) was 6.5±4.1 in participants with LBP, and 5.2±3.4 in participants without (p<0.0001). In multivariate analyses, LBP was significantly associated with increased WOMAC knee pain score (β[SE]=1.00[0.21]; p=<.0001). However, pain in all other individual musculoskeletal locations demonstrated similar associations with knee pain score. In models including all pain locations simultaneously, only LBP (β[SE]=0.65[0.21];p=.002), ipsilateral elbow pain (0.98 [0.40]; p=.02), and ipsilateral foot pain (1.03[0.45]; p=.02) were significantly associated with knee pain score. Having more than one pain location was associated with greater WOMAC knee pain; this relationship was strongest for individuals having four (β[SE]= 1.83[0.42]; p<0.0001), or five or more pain locations (1.86[0.36]; p<0.0001).
Conclusions
LBP, foot pain, and elbow pain are significantly associated with WOMAC knee pain score, as are a higher total number of pain locations. This may have implications for clinical trial planning.
Osteoarthritis (OA) of the knee is the leading cause of disability in the United States1. Low back pain (LBP) is the most common cause of time lost from work among individuals under the age of 45 years and the third most common cause among individuals between 45 and 65 years of age.2-3 The coexistence of LBP in individuals with knee pain may predispose to symptom severity well beyond the situation of isolated knee pain. Indeed, cross-sectional studies of individuals with knee OA have suggested that concurrent LBP may be associated with greater knee OA symptoms.4-5 Wolfe reported an association between concurrent LBP and higher scores on the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC). However, this study used a version of the WOMAC applied ‘globally’ without referring to the knee joint specifically4. Longitudinal studies have demonstrated that the presence of pre-operative LBP is one of the most important factors associated with poor pain outcomes after total knee arthroplasty (TKA)6 and revision TKA7. This small body of evidence supports a relationship between concurrent LBP and knee-specific functional limitations, but the nature of this relationship, and other factors which may affect this relationship, remain unexplored. Pain in other musculoskeletal locations, including the hip and the foot, may also be associated with symptoms in the knee8-9.
There are multiple mechanisms by which pain in locations external to the knee may be associated with increased knee pain. First, LBP and other joint pains may directly cause increased knee pain due to the biomechanical interrelationship of joints in the kinetic chain10-11. Second, other pain locations may be associated with factors which themselves cause knee-related functional limitations and pain. For example, the presence of LBP may simply be a marker for individuals prone to pain states, who may intrinsically experience higher levels of knee-related symptoms12-13. In this situation, we might expect any musculoskeletal pain location to be associated with knee pain, regardless of proximity or biomechanical relationship to the knee joint. Furthermore, an increased number of pain locations may be a marker of widespread pain comorbidity and may be associated with still higher levels of knee-related symptoms14. Third, problems of pain attribution and localization may create a situation where an individual is unable to discriminate between different joint-specific sources of pain15-16. This may cause the WOMAC to include pain reporting external to the knee, and effectively act as a global measure even when asked in a joint-specific manner.
Understanding of the mechanisms by which LBP and other musculoskeletal pain are associated with knee pain may help to identify patients who are at risk for poor outcomes following TKA, and patients who may benefit from co-interventions to treat musculoskeletal pain in other locations. Furthermore, the effects of other musculoskeletal pain on knee pain may have implications for studies that use the WOMAC as an outcome but do not account for pain external to the knee.
This study seeks to evaluate the association between concurrent LBP, and other musculoskeletal pain comorbidity, and WOMAC knee pain, in a population of individuals with symptomatic knee OA. We hypothesized that LBP would be associated with higher knee-specific WOMAC pain scores. However, we expected that other musculoskeletal pain sites would also be similarly associated with higher knee-specific WOMAC pain scores.
Materials and Methods
Participants
Data used in the preparation of this article were obtained from the Osteoarthritis Initiative (OAI) database, which is available for public access at http://www.oai.ucsf.edu/ (#AllClinical00, V0.2.2). The Osteoarthritis Initiative (OAI) is a publicly available multi-center population-based observational cohort study of knee OA, which is comprised of three groups, the Progression subcohort (n=1389), the Incidence subcohort (n=3285), and the Non-exposed Control group (N=122). This study included individuals from the Progression subcohort of OAI, consisting of individuals age 45-79 with symptomatic tibiofemoral knee OA in at least one knee at baseline. Symptomatic tibiofemoral knee OA was defined by 1) participant report of frequent knee symptoms defined as ‘pain, aching or stiffness in and around the knee on most days’ for at least one month during the past 12 months, and 2) radiographic evidence of tibiofemoral knee OA defined as the presence of an OARSI atlas osteophyte grade I-III (equivalent to Kellgren and Lawrence grade ≥ 2) on a fixed flexion radiograph based on the individual clinic readings. Exclusion criteria included participant report of rheumatoid arthritis or inflammatory arthritis; having severe joint space narrowing in both knees or unilateral TKA and severe joint space narrowing in the contralateral knee; inability to undergo 3.0 Tesla MRI examination of the knee; a positive pregnancy test; inability to provide a blood sample; requiring ambulatory aids aside from the use of a single straight cane 50% of ambulation time or less; co-morbid conditions that might interfere with ability to participate in a study with a 4-year follow-up time; unlikelihood to reside in the clinic area for at least 3 years; current participation in a double-blind randomized controlled trial; and unwillingness to sign informed consent.
Demographic Information
All participants in the Progression cohort received a standard battery of questions during the initial OAI eligibility interview, the OAI screening visit or the OAI enrollment visit. This baseline assessment of participant demographics included participant self report of age, gender, race, ethnicity, employment status, yearly income, educational attainment, and health insurance status. Employment status was defined as being currently self-employed for pay at the enrollment visit. Participants who were on leave but expecting to return to work within six months were considered to be employed. Yearly income was identified as less than $50,000, or $50,000 or greater. Health insurance coverage status was identified as ‘currently having any kind of health coverage’.
Medical History and Mental Health Comorbidity
Medical comorbid conditions were assessed by a self-reported version of the Charlson Comorbidity Index. The Charlson Comorbidity Index is a commonly used measure of comorbidity burden that has demonstrated validity and reliability17-19. Smoking history was assessed by the question ‘Do you smoke cigarettes now?’. Depression was measured by participant self-report using the Center for Epidemiologic Studies Depression Scale (CES-D). The CES-D has been shown to have valid and reliable psychometric properties20-21. Anxiety was measured by item 9 of the Short Form 12 (SF-12), in which the participant answers the question ‘How much of the time during the past 4 weeks have you felt calm and peaceful?’, using a 5-point Likert scale ranging from ‘1’ (‘all of the time’) to ‘5’ (‘none of the time’)22. Higher scores indicate greater anxiety. Fatigue was measured by item 10 of the SF-12, using an identical 5-point Likert scale response to the question ‘How much of the time during the past 4 weeks did you have a lot of energy’. Higher scores indicate greater fatigue. The SF-12 is a widely used, validated, and reliable measure of health-related quality of life22. Although items 9 and 10 of the SF-12 have not been validated for item-specific use as measures of anxiety or fatigue, they have been used in this manner previously23.
Anthropometric Characteristics
Examination measures were obtained at the enrollment visit of the OAI. Body mass index was calculated as weight (kg) divided by the square of height (m2). Abdominal circumference was assessed with the participant standing, using a tape measure over bare skin24.
Radiographic Assessment of Knee OA
Fixed flexion knee radiographs for assessment of the each tibiofemoral joint was obtained at the enrollment visit for each participant using a Synaflexor. Knee radiographs were interpreted by readers at the OAI Clinical Centers who were specifically trained to assess the baseline fixed flexion knee radiographs using a classification based on the OARSI atlas grading system 25. A simulated Kellgren-Lawrence grade was used for assessment of joint space narrowing and osteophytosis26.
Musculoskeletal Pain Comorbidity
The presence of musculoskeletal pain was identified by participant self-report of back pain, neck pain, shoulder pain, elbow pain, wrist pain, hand pain, ankle pain, hip pain, knee pain, ankle pain, and foot pain. Back pain was defined by the question ‘During the past 30 days, have you had any back pain?’. Back pain was characterized by severity as ‘mild’, ‘moderate’, or ‘severe’ on self report. The location of back pain was identified by the patient on a pain diagram, which allowed the selection of pain in ‘upper back’, ‘middle back’, ‘lower back’, and ‘buttocks’. LBP was defined as pain in the lower back or buttocks as per recent consensus criteria on optimal definitions of LBP location27. Neck pain, shoulder pain, elbow pain, wrist pain, hand pain, knee pain, ankle pain, and foot pain were defined by the question, ‘During the past 30 days, which of these joints have had pain, aching, or stiffness on most days?’. The locations of pain were then identified by the participant on a pain diagram, which permitted the identification of site-specific pain locations on either side of the body. The presence of right hip pain or left hip pain was defined by the question ‘During the past 12 months, have you had any pain, aching, or stiffness in your hip?’. Laterality of pain was specified by the patient for all pain locations, with the exception of LBP and neck pain.
Outcomes
Participants completed separate knee-specific WOMACs for both the right and left knee at the enrollment visit. The WOMAC is a widely used outcome measure for lower extremity OA, and has demonstrated reliability and validity in the context of knee OA28. The WOMAC consists of three knee-specific subscales: the WOMAC function subscale, the WOMAC pain subscale, and the WOMAC stiffness subscale. Since the WOMAC subscales are known to have substantial intercorrelations, the knee-specific WOMAC pain score was used as the primary outcome for this study29-30.
Statistical Analysis
We began by characterizing the prevalence of each musculoskeletal pain location in the entire cohort. The potential confounding variables of participant demographics, medical and psychiatric comorbidity, anthropometric features, musculoskeletal pain, and WOMAC scores were compared between the subgroup of patients with and without LBP. Educational attainment was dichotomized as less than college level education vs. college level or higher educational attainment. To estimate the association between the presence of LBP and knee-specific WOMAC pain score, we used linear regression to examine the subgroup of the entire Progression cohort who had no missing values for any of the variables initially examined. In the first stage of the analysis, we constructed a model examining the relationship between the independent variables of demographic features, medical and psychiatric comorbidity, anthropometric characteristics and OA grading, using the dependent variable of WOMAC pain score for the most symptomatic (painful) knee. We initially included all variables in the model, and employed a backward selection algorithm with a significance threshold of p=0.10 for variable removal. The variables of age, gender, race, and ethnicity were believed to have particular clinical importance and were forced into the model. In the second stage of the analysis, we examined the effects of specific musculoskeletal pain locations added to the model created in the first stage of the analysis. For extremity pain locations, we examined locations ipsilateral and contralateral to the more symptomatic knee separately. For axial pain in the neck and back, we did not have specific information on laterality of pain. First, each pain location was added separately to the first stage model, and associations between the pain location and WOMAC pain score were observed. Contralateral knee pain was included as an independent variable, though ipsilateral knee pain was not. Second, all pain locations were added simultaneously to the variables included in the first stage model, and the independent effects of each pain location were observed. We then conducted a series of analyses to compare the proportion of WOMAC pain score variance explained by different combinations of independent variables. We first removed from the full model all pain locations that were not significantly and independently associated with WOMAC pain score (p<0.05). We next constructed a model incorporating all pain locations and variables from the first stage model, but using three different levels of LBP severity: mild, moderate, and severe. Last, we constructed a model including the total number of pain comorbidities, irrespective of location, using the categories of 0, 1, 2, 3, 4, and 5 or more pain locations.
Results
The study sample was 57.1% female, with a mean ± SD age of 61.4 ± 9.1 years, and a BMI of 30.2 ± 4.9. The sample was 70.1 % white, 26.8 % black, <1.0 % Asian, and 2.2 % of other race. 1.5% reported Hispanic or Latino ethnicity. 51.5% of the sample had received a college education, 54.9 % of reported an annual income of more than $50,000, 95.5% had health insurance, and 59.6% were currently employed.
Table 1 demonstrates the prevalence of musculoskeletal pain by location in the study sample. The majority of participants (57.4%) reported having LBP in the past 30 days, while 20% reported neck pain. The prevalence of extremity pain at a joint on either side of the body that was present for more than half the days during the past 30 days was highest in the knee, and lowest in the elbow. Not all participants reported pain in a knee due to the fact that the definition of knee pain used to identify symptomatic knee OA in the parent study was more broad (symptoms during the past 12 months) than the definition used in this subanalysis (symptoms during the past 30 days). Hip aching or stiffness during the past 12 months was present in 59.6% of participants. A comparison of pain laterality for each joint demonstrates that a substantial percentage of individuals with pain at one joint also experienced pain in the same joint on the opposite side of the body.
Table 1. Prevalence of Musculoskeletal Pain by Location*.
| Pain Location | Either side | Right | Left |
|---|---|---|---|
| Spine Pain | |||
| Low back pain† | 798 (57.4%) | - | - |
| Neck Pain | 278 (20.0%) | - | - |
| Extremity Pain | |||
| Shoulder | 387 (27.8%) | 285 (20.5%) | 246 (17.7%) |
| Elbow | 172 (12.4%) | 110 (7.9%) | 112 (8.1%) |
| Wrist | 214 (15.4%) | 167 (12.0%) | 152 (10.9%) |
| Hand | 469 (33.7%) | 397 (28.6%) | 366 (26.3%) |
| Hip‡ | 828 (59.6%) | 670 (48.3%) | 561 (40.4%) |
| Knee | 1216 (87.9%) | 923 (66.7%) | 879 (63.5%) |
| Ankle | 219 (15.8%) | 175 (12.6%) | 154 (11.1%) |
| Foot | 230 (16.6%) | 192 (13.8%) | 175 (12.6%) |
Pain, aching or stiffness on more than half the days in past 30 days
Any low back or buttock pain in past 30 days
Any pain, aching or stiffness, past 12 months
Table 2 demonstrates the characteristics of individuals in the Progression cohort with respect to LBP status. Age, race, and ethnicity were comparable in participants with and without LBP. Participants with LBP were less likely to be male, be a college graduate, be currently employed, and have an annual income > $50,000. BMI and abdominal circumference were slightly higher in participants with LBP. Participants with LBP had higher scores on measures of psychiatric comorbidity, a higher prevalence of depression, greater overall comorbidity burden as reflected by the Charlson Comorbidity Index, and a higher prevalence of specific conditions likely to impact lower extremity function, including peripheral vascular disease, diabetes, and past stroke. Participants with LBP had higher frequencies of self-reported pain at every non-LBP pain location. Grade 2 and 3 knee OA was more common in participants with LBP. All WOMAC subscales were higher in participants with LBP (representing greater knee-related symptoms), with an average WOMAC pain score of 6.5 ± 4.1 in participants with LBP, and 5.2 ± 3.8 in participants without LBP.
Table 2. Relationships between LBP status and baseline characteristics in the progression cohort of the Osteoarthritis Initiative.
| Variable | + LBP (n=798) | No LBP (n=592) | p value |
|---|---|---|---|
| Demographics | |||
| Age | 61.2 ± 9.0 | 61.6 ± 9.3 | 0.52 |
| Female Gender | 476 (59.7%) | 317 (53.6%) | 0.02 |
| Race | |||
| White | 559 (70.1%) | 415 (70.1%) | 0.96 |
| Black | 214 (26.9%) | 158 (26.7%) | |
| Asian | 6 (0.8%) | 6 (1.0%) | |
| Other non-white | 18 (2.3%) | 13 (2.2%) | |
| Hispanic | 11 (1.4%) | 10 (1.7%) | 0.64 |
| Education – College Graduate | 383 (48.6%) | 324 (55.4%) | 0.01 |
| Health Insurance | 745 (94.8%) | 561 (96.4%) | 0.16 |
| Yearly Income-> $50,000 | 395 (52.5%) | 327 (58.2%) | 0.04 |
| Currently Employed | 460 (57.6%) | 369 (62.3%) | 0.08 |
| Anthropometric Features | |||
| Body Mass Index | 30.4 ± 5.1 | 29.6 ± 4.6 | 0.10 |
| Abdominal Circumference (cm) | 106.3 ± 13.3 | 104.9 ± 12.5 | 0.04 |
| Medical and Psychosocial Comorbidities | |||
| Depression (CES-D ≥16) | 121 (15.4%) | 68 (11.7%) | 0.05 |
| Anxiety1 | 2.31 ± 0.80 | 2.21 ± 0.76 | 0.02 |
| Fatigue1 | 2.75 ± 0.89 | 2.59 ± 0.89 | 0.001 |
| Smoking (current) | 73 (9.2%) | 31 (5.2%) | 0.006 |
| History of stroke | 26 (3.3%) | 19 (3.3%) | 0.003 |
| Diabetes | 90 (11.7%) | 61 (10.6%) | 0.54 |
| PAD (s/p bypass surgery) | 13 (1.7%) | 7 (1.2%) | 0.50 |
| Charlson comorbidity index | |||
| 0 | 516 (66.1%) | 433 (74.7%) | 0.001 |
| ≥1 | 265 (33.9%) | 147 (25.%) | |
| Musculoskeletal Pain Comorbidities* | |||
| Neck pain | 208 (26.1%) | 70 (11.8%) | < .0001 |
| Shoulder pain | 278 (34.8%) | 109 (18.4%) | < .0001 |
| Elbow pain | 128 (16.0%) | 44 (7.4%) | < .0001 |
| Wrist pain | 157 (19.7%) | 57 (9.6%) | < .0001 |
| Hand pain | 299 (37.5%) | 170 (28.7%) | < .0001 |
| Hip pain† | 553 (69.3%) | 275 (46.5%) | < .0001 |
| Knee pain | 713 (89.6%) | 503 (85.5%) | 0.02 |
| Ankle pain | 161 (20.2%) | 58 (9.8%) | < .0001 |
| Foot pain | 157 (19.7%) | 73 (12.3%) | < .0003 |
| Knee Osteoarthritis and Knee Symptoms‡ | |||
| Radiographic tibiofemoral knee OA2 | |||
| Grade 2 | 233 (29.5%) | 162 (27.6%) | |
| Grade 3 | 326 (41.2%) | 227 (38.7%) | 0.05 |
| Grade 4 | 160 (20.2%) | 157 (26.8%) | |
| WOMAC pain score | 6.5 ± 4.1 | 5.2 ± 3.8 | < .0001 |
| WOMAC total score | 29.5 ± 18.1 | 24.4 ± 17.0 | < .0001 |
Pain, aching or stiffness on more than half the days in past 30 days
Any pain, aching or stiffness, past 12 months
Values listed are for most painful knee as evidenced by WOMAC pain score
Anxiety and fatigue scales were taken from the SF-12, which grades these symptoms on a 5 point Likert-type scale from ‘1’- ‘5’ with ‘1’ as no symptoms, and ‘5’ as severe symptoms
Simulated Kellgren-Lawrence grade
In the first stage of the linear regression analysis, we used the outcome of most symptomatic WOMAC pain score and considered all variables from Table 2 (except for pain locations) for inclusion in the model. In order to facilitate comparisons between models, we limited the sample to participants with no missing values (n=1219). Age, gender, race, ethnicity, college education, income >$50,000, BMI, abdominal circumference, depression, fatigue, current smoking, Charlson comorbidity score ≥1, and OA grade were included in the final model after the backwards selection algorithm. The first stage regression model had a total variance (R2) of 0.223, and adjusted R2 of 0.213.
In the second stage of the analysis, we added each pain location separately as a independent variable to the first stage model. Table 3 depicts the results of linear regression analyses including pain locations while adjusting for all factors from the first stage model. The leftmost columns demonstrate the association between each individual pain location and WOMAC pain score when only one pain location was added to the model. The rightmost columns allow the reader to see the effects of individual pain locations in the full model which includes all pain locations added to the model simultaneously. Although all single pain locations demonstrated statistically significant associations with WOMAC pain score and comparable β coefficients when added to the model individually, most pain locations were not significantly and independently associated with WOMAC pain score when all pain locations were included in the model simultaneously. When adjusting for all pain locations, only LBP, ipsilateral elbow pain, and ipsilateral foot pain were significantly associated with WOMAC pain score (p<0.05). The second stage linear regression including all pain locations had a variance of R2 of 0.271, and adjusted R2 of 0.252. A reduced model including only the statistically significant pain locations of LBP, ipsilateral elbow pain, and ipsilateral foot pain had a variance of R2 of 0.258, and adjusted R2 of 0.247. A statistical comparison to detect a difference in model fit between the reduced model and the all-locations model approached significance p=.08. In secondary analyses intended to account for individuals with substantial pain in both knees, and factors specific to laterality of knee pain, we examined associations between all variables included in the second stage model, and the outcome of unilateral WOMAC pain score (right and left). LBP was independently and significantly associated with both right (β[SE]= 0.57 [0.20]; p=0.005) and left-sided (β[SE]= 0.67 [0.22]; p=0.0001) WOMAC knee pain score, however, ipsilateral foot pain and elbow pain were not. Ipsilateral ankle pain was the only other pain location that was associated with both right (β[SE]= 1.10 [0.39]; p=0.005) and left-sided (β[SE]= 1.07 [0.45]; p=0.02)WOMAC knee pain score.
Table 3. Associations between other musculoskeletal pain locations and WOMAC knee pain*.
| For individual musculoskeletal pain locations | Including all musculoskeletal pain locations | |||
|---|---|---|---|---|
| Pain Location† | β (SE) | p value | β (SE) | p value |
| Low Back Pain‡ | 1.00 (0.21) | <.0001 | 0.65 (0.21) | 0.002 |
| Neck Pain | 0.88 (0.26) | 0.0007 | 0.38 (0.27) | 0.16 |
| Shoulder pain | ||||
| Ipsilateral | 0.78 (0.26) | 0.002 | 0.08 (0.29) | 0.79 |
| Contralateral | 0.84 (0.27) | 0.002 | 0.21 (0.48) | 0.48 |
| Elbow pain | ||||
| Ipsilateral | 1.59 (0.36) | <0.0001 | 0.98 (0.40) | 0.02 |
| Contralateral | 1.14 (0.40) | 0.005 | -0.24 (0.45) | 0.60 |
| Wrist pain | ||||
| Ipsilateral | 1.01 (0.32) | 0.001 | -0.02 (0.41) | 0.97 |
| Contralateral | 1.28 (0.34) | 0.0002 | 0.38 (0.44) | 0.39 |
| Hand pain | ||||
| Ipsilateral | 0.70 (0.23) | 0.002 | 0.09 (0.32) | 0.77 |
| Contralateral | 0.77 (0.24) | 0.001 | 0.14 (0.33) | 0.68 |
| Hip pain1 | ||||
| Ipsilateral | 0.57 (0.21) | 0.006 | 0.09 (0.22) | 0.69 |
| Contralateral | 0.76 (0.21) | 0.0003 | 0.38 (0.23) | 0.09 |
| Knee pain | ||||
| Contralateral | 0.59 (0.21) | 0.005 | 0.36 (0.21) | 0.09 |
| Ankle pain | ||||
| Ipsilateral | 1.25 (0.31) | <0.0001 | 0.06 (0.41) | 0.89 |
| Contralateral | 1.64 (0.35) | <0.0001 | 0.71 (0.46) | 0.13 |
| Foot pain | ||||
| Ipsilateral | 1.46 (0.31) | <0.0001 | 1.03 (0.45) | 0.02 |
| Contralateral | 1.16 (0.31) | 0.0002 | -0.32 (0.45) | 0.48 |
WOMAC pain score for the most symptomatic knee, adjusting for age, gender, race, ethnicity, education, income, BMI, abdominal circumference, depression, fatigue, smoking, comorbidity burden, and OA grade
Pain, aching or stiffness on more than half the days in past 30 days
Any low back or buttock pain in past 30 days
Any pain, aching or stiffness, past 12 months
Table 4 depicts the results of linear regression analysis with the full model including all pain locations, but including stratification by LBP severity, using ‘no LBP’ as the reference group. When controlling for all other factors including other musculoskeletal pain, mild LBP had no effect on WOMAC pain score. Moderate LBP and severe LBP, however, demonstrated significant and independent associations with WOMAC pain score, with β coefficients of 1.08 and 1.93 respectively. The linear regression including all pain locations and LBP severity had a total variance of R2 of 0.288, and adjusted R2 of 0.268.
Table 4. Associations between low back pain severity and WOMAC knee pain score†.
| Pain Location | Prevalence | β (SE) | p value |
|---|---|---|---|
| Low back pain | |||
| None | 592 (42.6%) | (reference) | |
| Mild | 341 (24.5%) | -0.06 (0.25) | 0.79 |
| Moderate | 379 (27.3%) | 1.21 (0.25) | <0.0001 |
| Severe | 76 (5.5%) | 2.09 (0.52) | <0.0001 |
WOMAC pain score for the most symptomatic knee, adjusting for age, gender, race, ethnicity, education, income, BMI, abdominal circumference, depression, fatigue, smoking, comorbidity burden, OA grade, and all other pain locations
We conducted further analyses in order to examine whether specific pain locations are less important in explaining the variance in WOMAC pain score than total number of pain locations. In a linear regression model including all variables from the first stage model, and including total number of pain locations in the categories of 0, 1, 2, 3, 4, and 5 or more pain locations, we found that having 2, 4, or 5 or more pain locations was associated with higher WOMAC pain scores. Table 5 depicts the independent association between number of pain locations and the outcome of WOMAC pain score in a linear regression model including all variables from the first stage model, and including total number of pain locations in the categories of 0, 1, 2, 3, 4, and ≥ 5 pain locations. We found that having 2 or 3 pain locations showed an association with higher average WOMAC pain scores of borderline statistical significance (β[SE]=0.78 and 0.71, respectively), while having 4 or ≥5 pain locations was associated with substantially higher average WOMAC pain scores (β[SE]=1.83 and 1.86, respectively). The model including continuous number pain locations had a total variance of R2 of 0.250, and adjusted R2 of 0.238.
Table 5. Associations between number of pain locations and WOMAC knee pain score†.
| Number of Pain Locations | Prevalence | β (SE) | p value |
|---|---|---|---|
| 0 | 159 (11.5%) | (reference) | |
| 1 | 217 (15.6%) | 0.38 (0.39) | 0.32 |
| 2 | 220 (15.9%) | 0.78 (0.38) | 0.04 |
| 3 | 227 (16.4%) | 0.71 (0.38) | 0.06 |
| 4 | 151 (10.9%) | 1.83 (0.42) | <0.0001 |
| ≥5 | 414 (29.8%) | 1.86 (0.36) | <0.0001 |
WOMAC pain score for the most symptomatic knee, adjusting for age, gender, race, ethnicity, education, income, BMI, abdominal circumference, depression, fatigue, smoking, comorbidity burden, and OA grade.
Discussion
In this study of individuals with symptomatic knee OA, we found that any single musculoskeletal pain location external to the knee was associated with higher WOMAC knee pain scores. However, when all pain locations were taken into account, only the associations of LBP, ipsilateral foot pain, and ipsilateral elbow pain were significant. Although mild LBP was not associated with WOMAC knee pain score, moderate and severe LBP were each associated with substantially higher WOMAC knee pain scores. In contrast, regression models accounting for number of pain locations found an association between having 2, 4, or 5 or more pain locations at any site, and higher WOMAC knee pain score. These models explained a comparable, though slightly smaller, amount of the variance in WOMAC knee pain scores as compared to a parsimonious model including only the pain locations of LBP, ipsilateral elbow pain, and ipsilateral foot pain.
Our finding that LBP and ipsilateral foot pain are significantly and independently associated with higher WOMAC knee pain scores supports the commonsense clinical view that pain and function in any joint affects nearby joints both ‘above’ and ‘below’ in the kinetic chain, but would not be directly related to pain in a more distant location. Indeed, the idea that low back pain is biomechanically linked to knee pain via the so-called ‘knee-spine syndrome’ has been proposed in the spine literature10-11 It is noteworthy that pain at locations immediately adjacent to the knee (i.e. hip and ankle) was not independently associated with knee pain intensity, whereas pain localized one joint removed from the knee in the kinetic chain (i.e. low back and foot) was independently associated with knee pain intensity. This lack of an association between immediately adjacent joint pain (at the hip and ankle) and knee pain in this study suggests that individuals are able to accurately localize knee pain to a specific joint. However, true knee pain may also be confused with referred pain from the spine such as in the case of nerve root impingement and radiculopathy. A prior study by Wood has drawn attention to radicular pain from spinal pathology as a possible cause of non-articular knee pain31. Foot pain has also been found to be associated with greater knee symptoms in one prior study which did not account for non-lower extremity pain9.
In contrast to the associations with foot and low back pain, our finding that ipsilateral elbow pain is related to knee pain severity is neither intuitive, nor easy to explain from either a biomechanical or biological perspective. Although use of an assistive device causing upper extremity overuse seemed a possible mechanism, adjustment for use of an assistive device during walking performance tasks in the baseline OAI assessment did not materially change the magnitude or significance of the relationship between ipsilateral elbow pain and ipsilateral knee pain (data not shown). Given the many pain locations examined in this study, this association may have been a consequence of multiple statistical comparisons.
The prevalence of LBP in our study sample was 57.4%. This estimate is quite similar to the prevalence of 54.6% reported by Wolfe in an outpatient rheumatology clinic population of patients with clinical knee OA5. This extraordinarily high concomitant frequency is not well recognized, nor is the finding of LBP commonly documented in the clinic. Our findings provide a potential explanation for why prior studies have found that preoperative LBP is a risk factor for poor outcomes following joint replacement6-7. Furthermore, our findings suggest that severity of LBP is an important correlate of increased knee symptoms. When accounting for LBP intensity, severe LBP was associated with a 2 point increase in knee specific WOMAC pain score, an amount that is comparable to previously reported relative thresholds for minimal clinically important difference in WOMAC pain score following rehabilitation interventions.32 Conversely, mild LBP was not associated with WOMAC. These findings build upon evidence from the geriatric mobility literature, which suggest associations between LBP severity and function in older adults33-34.
Our finding that any single pain location- regardless of proximity or biomechanical relatedness to the symptomatic knee- is associated with higher WOMAC pain scores, and that a greater number of pain locations is associated with still higher WOMAC pain scores, provides further insights into the relationship between concurrent musculoskeletal pain burden and WOMAC pain score. The presence of pain in multiple locations may identify individuals with a greater propensity to pain states12-13. Croft et al. reported that pain elsewhere was common in individuals with knee pain, and individuals with other musculoskeletal pain reported more severe knee pain8. Peat et al. reported that a higher number of pain locations in the hip and foot was associated with greater pain as measured by the validated WOMAC scale applied specifically to the knee, highlighting the importance of concurrent lower extremity pain in particular.9. In contrast, our analysis accounts for all common sites of concurrent musculoskeletal pain (with the exception of headache/facial pain), and demonstrates that a model limited to specific pain locations (LBP, foot pain, and elbow pain) explains only a slightly greater proportion of the variance in WOMAC knee pain score than a model including total pain comorbidity (i.e., number of pain locations). Total pain comorbidity and specific pain locations may therefore represent important separate- but interrelated- factors associated with higher WOMAC pain score. That is, although certain pain locations are more strongly associated with higher WOMAC pain score, progressively higher pain comorbidity burden irrespective of location captures information that is also related to higher WOMAC pain score. Accounting for pain intensity as part of the assessment of musculoskeletal pain comorbidity may reveal even stronger associations with WOMAC pain scores, and is an area worthy of future study.
Taken together, our findings suggest that pain external to the knee may exert small but clinically significant effects on WOMAC pain score, even when the WOMAC is applied in a knee-specific manner. Given that success vs. failure in clinical trials is often decided by a difference of one or two points on the WOMAC pain score, stratification by musculoskeletal pain comorbidity may be a factor worth considering in trial design. Furthermore, the demonstrated relationship of non-knee pain to knee-specific pain suggest that co-intervention for musculoskeletal pain comorbidity may augment treatment for the knee and improve overall knee-related outcomes. Future studies of this or similar questions will benefit from a longitudinal design, and may be incorporated into existing cohort studies.
This study does have limitations. First, subjects in our study represent a sample of patients with knee osteoarthritis. Although we believe this sample to be representative of patients with symptomatic knee OA, in theory these findings may not be generalizable to other populations. Second, although the case definitions for pain were uniform for most pain locations (‘aching or stiffness on more than half the days in past 30 days’), a different definition was used for LBP (‘any low back or buttock pain in past 30 days’). This may explain both the higher prevalence of LBP in our sample, and the effect of moderate or severe LBP- but not mild LBP- on WOMAC pain scores. It is possible that using a more permissive definition for LBP may have contributed to our findings. However, we believe the results of stratification by LBP severity show that, if anything, the more permissive definition would have biased towards the null. Third, the assessment of pain locations used in this study may have led to recall bias. Although such bias would be expected to affect measurement of pain in all locations equally, hip pain was an exception, and would have been most susceptible to differential bias. Last, multiple statistical comparisons were made in this analysis. Our intent was to identify important relationships, and pre-hoc statistical adjustments to account for multiple comparisons were not planned or performed.
The results of this study draw further attention to the fact that symptomatic knee OA rarely occurs in isolation. Pain in the low back and in the ipsilateral foot and elbow may be associated with greater knee pain. In addition, total number of pain comorbidities also appears to be associated with more severe symptoms. Future studies are needed to determine whether treatment of musculoskeletal pain comorbidity may improve the outcomes of treatment for knee OA.
Acknowledgments
We would like to thank the participants, Principal Investigators (Michael Nevitt, Kent Kwoh, Charles B. Eaton, Rebecca Jackson, Marc Hochberg, Joan Bathon), Co-Investigators and staff of the Osteoarthritis Initiative. The OAI is a public-private partnership comprised of five contracts (N01-AR-2-2258; N01-AR-2-2259; N01-AR-2-2260; N01-AR-2-2261; N01-AR-2-2262) funded by the National Institutes of Health, a branch of the Department of Health and Human Services, and conducted by the OAI Study Investigators. Private funding partners include Merck Research Laboratories; Novartis Pharmaceuticals Corporation, GlaxoSmithKline; and Pfizer, Inc. Private sector funding for the OAI is managed by the Foundation for the National Institutes of Health. This manuscript has received the approval of the OAI Publications Committee based on a review of its scientific content and data interpretation.
Funding sources: The OAI is a public-private partnership comprised of five contracts (N01-AR-2-2258; N01-AR-2-2259; N01-AR-2-2260; N01-AR-2-2261; N01-AR-2-2262) funded by the National Institutes of Health, a branch of the Department of Health and Human Services, and conducted by the OAI Study Investigators. Private funding partners include Merck Research Laboratories; Novartis Pharmaceuticals Corporation, GlaxoSmithKline; and Pfizer, Inc. Private sector funding for the OAI is managed by the Foundation for the National Institutes of Health. Drs. Suri and Morgenroth are funded by the Rehabilitation Medicine Scientist Training Program (RMSTP) and the National Institutes of Health (K12 HD 01097). Dr. Hunter is funded by an ARC Future Fellowship.
Footnotes
Conflict of interest statement: None of the authors have any conflict of interest regarding the contents of this article.
References
- 1.Dunlop DD, Manheim LM, Yelin EH, Song J, Chang RW. The costs of arthritis. Arthritis Rheum. 2003;49:101–13. doi: 10.1002/art.10913. [DOI] [PubMed] [Google Scholar]
- 2.Cunningham LS, Kelsey JL. Epidemiology of musculoskeletal impairments and associated disability. American journal of public health. 1984;74:574–9. doi: 10.2105/ajph.74.6.574. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Kelsey JL, White AA., 3rd Epidemiology and impact of low-back pain. Spine (Phila Pa 1976) 1980;5:133–42. doi: 10.1097/00007632-198003000-00007. [DOI] [PubMed] [Google Scholar]
- 4.Wolfe F. Determinants of WOMAC function, pain and stiffness scores: evidence for the role of low back pain, symptom counts, fatigue and depression in osteoarthritis, rheumatoid arthritis and fibromyalgia. Rheumatology (Oxford) 1999;38:355–61. doi: 10.1093/rheumatology/38.4.355. [DOI] [PubMed] [Google Scholar]
- 5.Wolfe F, Hawley DJ, Peloso PM, Wilson K, Anderson J. Back pain in osteoarthritis of the knee. Arthritis Care Res. 1996;9:376–83. doi: 10.1002/1529-0131(199610)9:5<376::aid-anr1790090506>3.0.co;2-1. [DOI] [PubMed] [Google Scholar]
- 6.Escobar A, Quintana JM, Bilbao A, et al. Effect of patient characteristics on reported outcomes after total knee replacement. Rheumatology (Oxford) 2007;46:112–9. doi: 10.1093/rheumatology/kel184. [DOI] [PubMed] [Google Scholar]
- 7.Novicoff WM, Rion D, Mihalko WM, Saleh KJ. Does Concomitant Low Back Pain Affect Revision Total Knee Arthroplasty Outcomes? Clin Orthop Relat Res. 2009 doi: 10.1007/s11999-009-0882-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Croft P, Jordan K, Jinks C. “Pain elsewhere” and the impact of knee pain in older people. Arthritis Rheum. 2005;52:2350–4. doi: 10.1002/art.21218. [DOI] [PubMed] [Google Scholar]
- 9.Peat G, Thomas E, Wilkie R, Croft P. Multiple joint pain and lower extremity disability in middle and old age. Disabil Rehabil. 2006;28:1543–9. doi: 10.1080/09638280600646250. [DOI] [PubMed] [Google Scholar]
- 10.Murata Y, Takahashi K, Yamagata M, Hanaoka E, Moriya H. The knee-spine syndrome. Association between lumbar lordosis and extension of the knee. J Bone Joint Surg Br. 2003;85:95–9. doi: 10.1302/0301-620x.85b1.13389. [DOI] [PubMed] [Google Scholar]
- 11.Tsuji T, Matsuyama Y, Goto M, et al. Knee-spine syndrome: correlation between sacral inclination and patellofemoral joint pain. J Orthop Sci. 2002;7:519–23. doi: 10.1007/s007760200092. [DOI] [PubMed] [Google Scholar]
- 12.Natvig B, Bruusgaard D, Eriksen W. Localized low back pain and low back pain as part of widespread musculoskeletal pain: two different disorders? A cross-sectional population study. J Rehabil Med. 2001;33:21–5. doi: 10.1080/165019701300006498. [DOI] [PubMed] [Google Scholar]
- 13.Natvig B, Eriksen W, Bruusgaard D. Low back pain as a predictor of long-term work disability. Scand J Public Health. 2002;30:288–92. doi: 10.1080/14034940210133951. [DOI] [PubMed] [Google Scholar]
- 14.Kamaleri Y, Natvig B, Ihlebaek CM, Bruusgaard D. Localized or widespread musculoskeletal pain: does it matter? Pain. 2008;138:41–6. doi: 10.1016/j.pain.2007.11.002. [DOI] [PubMed] [Google Scholar]
- 15.Dawson J, Fitzpatrick R, Murray D, Carr A. The problem of ‘noise’ in monitoring patient-based outcomes: generic, disease-specific and site-specific instruments for total hip replacement. J Health Serv Res Policy. 1996;1:224–31. doi: 10.1177/135581969600100408. [DOI] [PubMed] [Google Scholar]
- 16.Johanson NA, Liang MH, Daltroy L, Rudicel S, Richmond J. J Bone Joint Surg Am. 86-A. 2004. American Academy of Orthopaedic Surgeons lower limb outcomes assessment instruments. Reliability, validity, and sensitivity to change; pp. 902–9. [DOI] [PubMed] [Google Scholar]
- 17.Kieszak SM, Flanders WD, Kosinski AS, Shipp CC, Karp H. A comparison of the Charlson comorbidity index derived from medical record data and administrative billing data. Journal of clinical epidemiology. 1999;52:137–42. doi: 10.1016/s0895-4356(98)00154-1. [DOI] [PubMed] [Google Scholar]
- 18.Poses RM, McClish DK, Smith WR, Bekes C, Scott WE. Prediction of survival of critically ill patients by admission comorbidity. Journal of clinical epidemiology. 1996;49:743–7. doi: 10.1016/0895-4356(96)00021-2. [DOI] [PubMed] [Google Scholar]
- 19.Rochon PA, Katz JN, Morrow LA, et al. Comorbid illness is associated with survival and length of hospital stay in patients with chronic disability. A prospective comparison of three comorbidity indices. Medical care. 1996;34:1093–101. doi: 10.1097/00005650-199611000-00004. [DOI] [PubMed] [Google Scholar]
- 20.Kessler RC, Foster CL, Saunders WB, Stang PE. Social consequences of psychiatric disorders, I: Educational attainment. The American journal of psychiatry. 1995;152:1026–32. doi: 10.1176/ajp.152.7.1026. [DOI] [PubMed] [Google Scholar]
- 21.Rushton JL, Forcier M, Schectman RM. Epidemiology of depressive symptoms in the National Longitudinal Study of Adolescent Health. Journal of the American Academy of Child and Adolescent Psychiatry. 2002;41:199–205. doi: 10.1097/00004583-200202000-00014. [DOI] [PubMed] [Google Scholar]
- 22.Hurst NP, Ruta DA, Kind P. Comparison of the MOS short form-12 (SF12) health status questionnaire with the SF36 in patients with rheumatoid arthritis. British journal of rheumatology. 1998;37:862–9. doi: 10.1093/rheumatology/37.8.862. [DOI] [PubMed] [Google Scholar]
- 23.Tiedemann A, Sherrington C, Lord SR. Physiological and psychological predictors of walking speed in older community-dwelling people. Gerontology. 2005;51:390–5. doi: 10.1159/000088703. [DOI] [PubMed] [Google Scholar]
- 24.Pearson AM, Lurie JD, Blood EA, et al. Spine patient outcomes research trial: radiographic predictors of clinical outcomes after operative or nonoperative treatment of degenerative spondylolisthesis. Spine (Phila Pa 1976) 2008;33:2759–66. doi: 10.1097/BRS.0b013e31818e2d8b. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Altman RD, Gold GE. Atlas of individual radiographic features in osteoarthritis, revised. Osteoarthritis Cartilage. 2007;15 A:A1–56. doi: 10.1016/j.joca.2006.11.009. [DOI] [PubMed] [Google Scholar]
- 26.Composite OA grading. [1/10/2009];2007 at http://www.oai.ucsf.edu/datarelease/DisplayCommentSASCode.asp?file=macros/CompositeOA.sas.
- 27.Dionne CE, Dunn KM, Croft PR, et al. A consensus approach toward the standardization of back pain definitions for use in prevalence studies. Spine (Phila Pa 1976) 2008;33 doi: 10.1097/BRS.0b013e31815e7f94. Online Supplement. [DOI] [PubMed] [Google Scholar]
- 28.Roos EM, Klassbo M, Lohmander LS. WOMAC osteoarthritis index. Reliability, validity, and responsiveness in patients with arthroscopically assessed osteoarthritis. Western Ontario and MacMaster Universities. Scandinavian journal of rheumatology. 1999;28:210–5. doi: 10.1080/03009749950155562. [DOI] [PubMed] [Google Scholar]
- 29.Suarez-Almazor ME, Souchek J, Kelly PA, et al. Ethnic variation in knee replacement: patient preferences or uninformed disparity? Archives of internal medicine. 2005;165:1117–24. doi: 10.1001/archinte.165.10.1117. [DOI] [PubMed] [Google Scholar]
- 30.Ryser L, Wright BD, Aeschlimann A, Mariacher-Gehler S, Stucki G. A new look at the Western Ontario and McMaster Universities Osteoarthritis Index using Rasch analysis. Arthritis Care Res. 1999;12:331–5. doi: 10.1002/1529-0131(199910)12:5<331::aid-art4>3.0.co;2-w. [DOI] [PubMed] [Google Scholar]
- 31.Wood LR, Peat G, Thomas E, Duncan R. The contribution of selected non-articular conditions to knee pain severity and associated disability in older adults. Osteoarthritis Cartilage. 2008;16:647–53. doi: 10.1016/j.joca.2007.10.007. [DOI] [PubMed] [Google Scholar]
- 32.Angst F, Aeschlimann A, Michel BA, Stucki G. Minimal clinically important rehabilitation effects in patients with osteoarthritis of the lower extremities. J Rheumatol. 2002;29:131–8. [PubMed] [Google Scholar]
- 33.Morone NE, Karp JF, Lynch CS, Bost JE, El Khoudary SR, Weiner DK. Impact of chronic musculoskeletal pathology on older adults: a study of differences between knee OA and low back pain. Pain Med. 2009;10:693–701. doi: 10.1111/j.1526-4637.2009.00565.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Weiner DK, Haggerty CL, Kritchevsky SB, et al. How does low back pain impact physical function in independent, well-functioning older adults? Evidence from the Health ABC Cohort and implications for the future. Pain Med. 2003;4:311–20. doi: 10.1111/j.1526-4637.2003.03042.x. [DOI] [PubMed] [Google Scholar]