Abstract
Background
Little is known about the temporal evolution of pain severity in persons with knee OA. We sought to describe the pain trajectory over 6 years in a cohort of subjects with radiographic, symptomatic knee OA.
Methods
We used data from the Osteoarthritis Initiative (OAI), a multi-center, longitudinal study of subjects with diagnosed radiographic evidence of knee OA. Pain was assessed at baseline and annually for 6 years. Our analysis cohort included subjects with symptomatic knee OA at baseline, defined as baseline Kellgren-Lawrence (KL) score ≥2 with WOMAC pain score >0. We used group-based trajectory modeling to identify distinct patterns of pain progression over 6-year follow-up. Factors examined included sex, race, education, comorbidities, age, body mass index (BMI), alignment, KL grade, and depression.
Results
We used data from 1,753 OAI participants with symptomatic knee OA. Mean baseline WOMAC pain score was 26.5 (0–100,100 = worst) with standard deviation 19. Group-based trajectory modeling identified 5 distinct pain trajectories; baseline pain scores for each ranged from 15 to 62. None of the trajectories exhibited substantial worsening. One fifth of subjects in the two trajectories with the greatest pain underwent total knee replacement over follow-up. Higher KL grade, obesity, depression, medical comorbidities, female sex, non-white race, lower education, and younger age were associated with trajectories characterized by greater pain.
Conclusion
We found that knee pain changes little, on average, over six years in most subjects. These observations suggest knee OA is characterized by persistent rather than inexorably worsening symptoms.
Keywords: osteoarthritis, pain, trajectories, group-based trajectory modeling, cohort study
Introduction
Symptomatic knee osteoarthritis (OA) has become a growing burden for patients and the broader American healthcare system, occurring in an estimated 6% of adults 30 years of age or older1 and in 13% of people age 60 and over2. Persons with OA of the lower extremities have lower quality of life compared to persons without OA3 and utilize more healthcare resources4,5. Knee pain is the primary reason that people with knee OA seek medical care6.
Pathologically, the disease is characterized by progressive hyaline articular cartilage damage coupled with changes in subchondral bone and other joint structures. In the course of these structural changes in the joint, affected persons may experience both symptomatic joint pain as well as functional disability1. Beyond replacing the joint surgically through arthroplasty, there are no treatments available to reverse the course of structural progression.
While associations between structural change, symptoms, and functional impairment are not well understood, both structural deterioration and symptoms are thought to gradually and consistently worsen over time7,8. Recent work challenges this traditional understanding. Felson et al. evaluated structural changes in knee OA over time by studying radiographic images from the Osteoarthritis Initiative (OAI). The study suggested that structural progression fits a pattern of inertia: knees that have begun progressing are likely to experience further worsening whereas knees that have been stable are likely to remain stable9. Peters et al. and Dieppe et al. each conducted longitudinal studies evaluating cohorts with knee OA over 7- and 8-years, respectively. While both study cohorts demonstrated overall declines in pain and function over time, the outcomes of individual subjects within these cohorts were heterogeneous with some subjects experiencing worsening and others improvement10,11.
These studies of symptomatic and structural progression suggest that persons with knee OA may have diverse disease trajectories over time; however, traditional approaches to longitudinal data analysis may not be suitable in the presence of heterogeneity leading to distinct outcome trajectories12. More work is needed to identify distinct trajectories in the natural history of pain for persons with knee OA; indeed, better understanding of these trajectories in persons affected by knee OA would offer important insights into clinical prognosis and would help inform treatment plans. This study seeks to describe the trajectory of OA-related pain over the course of six years by examining a large cohort of subjects with radiographic, symptomatic knee OA. A group-based trajectories approach has been designed to highlight the distinct pattern of outcomes.
Methods
Sample
We selected data from the OAI, a multi-center, longitudinal observational study of knee OA. Men and women ages 45–79 were enrolled at four centers between 2004 and 2006. Subjects were assessed annually; as of October 2013, clinical data were available through the 72 month visit. The data and additional study details are publicly available at http://oai.epi-ucsf.org. We selected all knees with radiographic knee OA at baseline, defined by Kellgren-Lawrence (KL) score of two or higher based on the central reading of a standardized fixed-flexion radiograph13. For subjects with radiographic OA in both knees, we selected the knee with the worst pain at baseline to create a database with one observation per subject. For subjects with radiographic OA in both knees and equal pain in each knee, we randomly selected a knee. We censored subjects that underwent primary total knee replacement (TKR) over the course of the study at the time of TKR; data was included up to and including the visit prior to TKR.
Primary Outcome
Our primary outcome was the pain subscale from the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC)14. The WOMAC pain scale contains 5 items rated on a Likert scale from 0 (no pain) to 4 (extreme pain). The items are summed to create a total score that was rescaled to 0–100, where higher scores indicate more severe pain. In describing baseline pain, thresholds of 15 and 40 were used to define categories of pain severity. Scores ≤ 15 were defined as mild pain, 15 to 40 as mild to moderate pain, and above 40 as moderate to severe pain. The lower threshold was selected to reflect the point at which subjects experience clinically relevant OA-related pain15,16. Our mid-point definition was based on literature describing pain scores of roughly 40 to 50 points as categorically moderate17,18.
Factors Associated with Pain
We examined the impact of baseline age, sex, BMI, comorbidities, KL grade, alignment, education, race, depression, and medication use on pain trajectories. Alignment was measured on full limb radiographs using a goniometer19. Comorbidites were summarized with the Charlson Comorbidity index20. Depression was defined as a Center for Epidemiologic Studies Depression Scale (CES-D) score > 1621. In multivariable models, BMI was classified as obese (> 30 kg/m2) vs. non-obese, race as white vs. non-white, education as college vs. no college, and comorbidities as any vs. none. Frequent medication use was defined as use of medication for knee pain more than half the days of the month in the past 12 months.
Statistical Methods
We used group-based trajectory modeling to identify groups of individuals following similar patterns of pain progression12. This method uses a multinomial modeling strategy to identify multiple trajectories, allowing for the detection of distinct outcome groups. Unlike growth curve modeling, which allows for individual variability around a mean population trend, group-based trajectory modeling allows for the possibility that there are distinct sub-groups within a population12. We used a censored normal model, which allows for clustering at the minimum and maximum and is useful for scales with a pre-specified range such as the WOMAC. We iteratively considered between 3 and 6 trajectories and allowed for up to a fourth-order polynomial in each trajectory. We used the Bayesian information criterion (BIC) to aid in selecting the optimal model. The BIC is a measure of model fit that balances improvements in model likelihood with the number of parameters estimated. Higher BIC values indicate better model fit; however, the BIC does not always clearly identify the optimal number of groups in group-based trajectory modeling12. Therefore, we also took into account group size and stability. We sought models with at least 50 participants in each group (approximately 3% of the sample) and models in which size and shape of each trajectory did not change substantially as we adjusted the order of polynomial.
We used posterior group-membership probabilities to assess model fit. For a specific individual, the posterior group-membership probability measures the likelihood of belonging to each trajectory group. For example, in a model with 3 trajectories, each individual has 3 posterior group membership probabilities that sum to 1 indicating the likelihood of belonging to each trajectory. Individuals are then assigned to the trajectory with the maximum posterior group-membership probability. Ideally, this probability is close to 1, suggesting only a small chance that the individual could belong to a different trajectory than the one assigned. Nagin suggests that the average posterior probability in each trajectory should be at least 0.712. We assessed heterogeneity within each trajectory by calculating the proportion of subjects demonstrating sustained worsening or improvement, which we defined as a change greater than the minimal clinically important difference (MCID) sustained at subsequent visits. The MCID was estimated on the WOMAC scale by Angst and colleagues as 11 points for worsening and 7.5 points for improvement15.
We used a three-stage approach to investigate the association between baseline covariates and the likelihood of membership in a specific trajectory. First, we selected the optimal model as described above. Then we used a multivariable multinomial logit model to examine the association between trajectory group assignment and covariates. In this step, we identified statistically significant predictors of trajectory group assignment. Finally, we re-estimated the trajectories by including the covariates statistically significantly associated with a higher likelihood of a specific pain trajectory in stage two, jointly estimating the trajectories and associations between the trajectories and covariates. This association between covariates and trajectory group membership was evaluated with the multivariable multinomial logit model. This allowed for the evaluation of one covariate’s influence on the probability of belonging to one trajectory over another comparative trajectory. This three-stage approach is necessary because the classification into each trajectory is based on the posterior group-membership probability. Uncertainty around these classifications must be taken into account when examining the association between group membership and covariates.
To investigate the association between pain and TKR, we plotted mean WOMAC pain by visit and TKR status. Following the methodology of Dodge et al., we created an indicator variable for whether or not each subject underwent TKR over the course of follow-up and a variable indicating at which visit the TKR was reported. We included the TKR information as a covariate and re-estimated the trajectories two ways: one model included the indicator for any TKR and one model included timing of the TKR visit22.
All analyses were conducted using SAS 9.3 (SAS Institute, Cary NC). We used SAS PROC TRAJ, a custom SAS procedure available for free download23,24.
Secondary Analysis
To assess more homogeneous disease groups, we evaluated pain trajectories separately for each baseline KL grade. We also analyzed trajectories for the WOMAC function scale. The WOMAC function scale contains 17 items rated on a Likert scale from 0 (no difficulty with item) to 4 (extreme difficulty), which were summed and rescaled to 0–100. Higher scores indicate greater functional impairment. Finally, we analyzed pain trajectories for all subjects with radiographic OA at baseline, including subjects with and without baseline pain.
Results
Sample
Our final sample was 1,753 subjects/knees with KL ≥2 and WOMAC pain >0 at baseline. The sample was 59% female with an average age of 62 (standard deviation (SD) 9). Forty-nine percent of the sample was obese with an average BMI of 30 (SD 5). Fifty-two percent had KL 2 radiographic severity, 34% KL 3, and 14% KL 4. The average WOMAC pain at baseline was 27 (SD 19). Forty-one percent of subjects had a WOMAC pain score ≤ 15, 40% were between 15 and 40, and 20% had a score greater than 40. The sample also reported functional impairment with an average WOMAC function score of 23 (SD 19) (Table 1). One hundred ninety-one subjects (11%) underwent primary TKR on the selected knee during the 72 months of follow-up. Among those who had TKR, 8% had surgery prior to the 12-month visit, 14% between 12 and 24 months, 21% between 24 and 36 months, 18% between 36 and 48 months, 20% between 48 and 60 months, and 19% between 60 and 72 months.
Table 1.
Baseline Characteristics.
| Age, mean (sd) | 62.2 (8.9) |
| n (%) Female | 1039 (59.3%) |
| BMI (mean (sd)) | 30.1 (4.9) |
| Comorbidities | |
| 0 | 1257 (72.5%) |
| 1 | 280 (16.1%) |
| 2+ | 197 (11.4%) |
| Race | |
| White or Caucasian | 1290 (73.6%) |
| Black or African American | 413 (23.6%) |
| Asian | 17 (1.0%) |
| Other Non-white | 33 (1.9%) |
| Education | |
| < High school | 80 (4.6%) |
| High school graduate | 717 (41.2%) |
| College graduate | 464 (26.6%) |
| Graduate degree | 481 (27.6%) |
| Depression (CES-D Score >= 16) | 205 (11.8%) |
| Kellgren and Lawrence | |
| 2 | 917 (52.3%) |
| 3 | 596 (34.0%) |
| 4 | 240 (13.7%) |
| Alignment | |
| No Malignment | 457 (26.2%) |
| Varus | 497 (28.5%) |
| Valgus | 792 (45.4%) |
| WOMAC Pain Score (scaled 0–100) (mean (sd)) | 26.5 (18.9) |
| 12M: WOMAC Pain Group | |
| Mild (<=15) | 712 (40.6%) |
| Moderate (<=40) | 696 (39.7%) |
| Severe (40+) | 345 (19.7%) |
| WOMAC Function Score (scaled 0–100) (mean (sd)) | 23.4 (18.5) |
| Number for whom either knee limited activities due to pain, aching or stiffness, past 30 days | 708 (40.5%) |
| Dropout | |
| No dropout | 1448 (82.6%) |
| After 60M | 58 (3.3%) |
| After 48M | 114 (6.5%) |
| After 36M | 37 (2.1%) |
| After 24M | 48 (2.7%) |
| After 12M | 48 (2.7%) |
| TKR | 191 (10.9%) |
| Prior to 12 Month | 15 |
| Between 12 and 24 Month | 27 |
| Between 24 and 36 Month | 40 |
| Between 36 and 48 Month | 35 |
| Between 48 and 60 Month | 38 |
| Between 60 and 72 Month | 36 |
| Medications | |
| Any medication for pain, aching or stiffness in the past 12 months | 1282 (73.1%) |
| Any medication for pain, aching or stiffness more than half the days of a month in the past 12 months | 709 (40.4%) |
| Any NSAID/Acetaminophen* | 793 (45.3%) |
| Nonprescription NSAIDS* | 452 (25.8%) |
| Prescription NSAIDS* | 145 (8.3%) |
| Tylenol* | 227 (13.0%) |
| COXIBS* | 172 (9.8%) |
| Strong prescription pain medications (e.g., narcotics)* | 51 (2.9%) |
| Doxycycline* | 5 (0.3%) |
| Any supplement** | 696 (39.7%) |
| Chondroitin sulfate** | 621 (35.6%) |
| Glucosamine** | 691 (39.5%) |
| Injections ** | 92 (5.3%) |
Use during more than half the days in the past 30 days
Use within the past 6 months
At baseline, 73% of the sample reported using medication for pain, aching or stiffness in the past 12 months, with 40% reporting using medication more than half the days of the month for the past 12 months. The most commonly used medications were nonprescription NSAIDS (26%), Tylenol (13%), and prescription NSAIDS (8%). Glucosamine (39%) and chondroitin sulfate (36%) were also popular treatments.
Eleven percent of subjects reported frequent medication use (using medications for knee pain, aching, or stiffness for more than half the days over the past 12 months) over the course of all follow-up visits, while 51% reported frequent medication use at some follow-up visits and not others. The remaining 38% did not report frequent medication use at any follow-up visit.
Group-based trajectory modeling
Our analysis identified 5 distinct pain trajectories: no pain, mild pain, low moderate pain, high moderate pain, and severe pain (Figure 1). The average posterior group-membership probability ranged from 0.80 for low moderate pain to 0.87 for severe pain, indicative of good model fit. None of the trajectories showed substantial worsening or improvement over time. Participants with moderate pain tended to stay in moderate pain while participants with more severe pain tended to stay in severe pain.
Figure 1. Group-Based Trajectories.
WOMAC pain is along the Y-axis and follow-up month is along the X-axis. Each line represents one distinct trajectory. The labels next to each trajectory indicate the percent of the population in that trajectory. The first trajectory contains 11% of the cohort, the second trajectory contains 34% of the cohort, the third trajectory contains 32% of the cohort, the fourth trajectory contains 17% of the cohort, and the fifth trajectory contains 6% of the cohort.
All of the trajectories suggested some improvement between baseline and month 12. For all subjects, the average improvement from baseline to month 12 was 6 points, while the subsequent yearly mean change between follow-up visits ranged from 0 points to a worsening of 2 points. This trend was not a simple matter of attrition; we saw similar findings when including only those subjects that did not dropout. As a sensitivity analysis, we restricted the analysis to visits occurring in months 12 through 72 (omitting baseline) and re-fit the trajectories. Group-based trajectory modeling again identified 5 distinct pain trajectories with results similar to the model including baseline data: none of the trajectories showing substantial worsening or improvement over the 12 to 72 month follow-up period (Supplementary Figure 1). The initial improvement seen after baseline was no longer observed when we restricted the analysis to months 12 to 72.
We observed some heterogeneity within each trajectory. Because of the observed trend of improvement after baseline, we evaluated sustained improvement and worsening from the month 12 visit. Overall, 85% of subjects experienced neither sustained worsening nor improvement. Analyses within trajectories showed that 94% of subjects in the no pain trajectory experienced neither worsening nor improvement, as compared with 88% of subjects in the mild pain trajectory, 80% of subjects in the low moderate pain trajectory, 80% of subjects in the high moderate pain trajectory, and 88% of subjects in the severe pain trajectory.
Impact of subject characteristics
In unadjusted analysis, higher KL grade, obesity, depression, comorbidities, female sex, non-white race, lower education, and younger age were associated with being in a trajectory characterized by greater pain (Table 2). With the exception of alignment, all baseline factors were significantly associated with trajectory in the multivariable model. These statistically significant associations between covariate and trajectory group remained when we examined bivariate associations between each covariate and trajectory group, rather than examining the associations in a multivariable fashion.
Table 2.
Baseline Characteristics by Trajectory (Unadjusted)
| None (n=187) | Mild (n=606) | Low moderate (n=563) | High moderate (n=292) | Severe (n=105) | |
|---|---|---|---|---|---|
| Age (mean (sd)) | 63.7 (8.5) | 62.1 (8.9) | 62.7 (8.8) | 61.5 (9.0) | 59.0 (8.7) |
|
| |||||
| Number Female (%) | 98 (52%) | 347 (57%) | 334 (59%) | 183 (63%) | 77 (73%) |
|
| |||||
| Mean BMI (sd) | 29.2 (4.3) | 29.3 (4.7) | 30.1 (4.7) | 31.1 (4.8) | 33.5 (5.7) |
|
| |||||
| Number Obese (%) | 79 (42%) | 254 (42%) | 279 (50%) | 164 (56%) | 76 (72%) |
|
| |||||
| Comorbidities | |||||
| 0 | 145 (78%) | 468 (78%) | 406 (73%) | 183 (63%) | 55 (55%) |
| 1 | 22 (12%) | 76 (13%) | 96 (17%) | 63 (22%) | 23 (23%) |
| 2+ | 19 (10%) | 55 (9%) | 57 (10%) | 44 (15%) | 22 (22%) |
|
| |||||
| Race | |||||
| White or Caucasian | 144 (77%) | 490 (81%) | 432 (77%) | 185 (63%) | 39 (37%) |
| Black or African American | 38 (20%) | 102 (17%) | 115 (20%) | 94 (32%) | 64 (61%) |
| Asian | 1 (1%) | 7 (1%) | 6 (1%) | 3 (1%) | 0 (0%) |
| Other Non-white | 4 (2%) | 7 (1%) | 10 (2%) | 10 (3%) | 2 (2%) |
|
| |||||
| Education | |||||
| < High school | 7 (4%) | 17 (3%) | 21 (4%) | 18 (6%) | 17 (17%) |
| High school graduate | 61 (33%) | 215 (36%) | 228 (41%) | 147 (51%) | 66 (64%) |
| College graduate | 60 (32%) | 156 (26%) | 167 (30%) | 69 (24%) | 12 (12%) |
| Graduate degree | 59 (32%) | 214 (36%) | 144 (26%) | 56 (19%) | 8 (8%) |
|
| |||||
| Depression (CES-D Score >= 16) | 6 (3%) | 43 (7%) | 64 (11%) | 55 (19%) | 37 (37%) |
|
| |||||
| Kellgren and Lawrence | |||||
| 2 | 127 (68%) | 385 (64%) | 242 (43%) | 121 (41%) | 42 (40%) |
| 3 | 54 (29%) | 171 (28%) | 214 (38%) | 120 (41%) | 37 (35%) |
| 4 | 6 (3%) | 50 (8%) | 107 (19%) | 51 (17%) | 26 (25%) |
|
| |||||
| Alignment | |||||
| Neither | 49 (26%) | 142 (23%) | 157 (28%) | 78 (27%) | 31 (30%) |
| Varus | 60 (32%) | 175 (29%) | 152 (27%) | 80 (28%) | 30 (29%) |
| Valgus | 78 (42%) | 288 (48%) | 252 (45%) | 131 (45%) | 43 (41%) |
|
| |||||
| Number for whom either knee limited activities due to pain, aching or stiffness, past 30 days | 39 (21%) | 201 (33%) | 224 (40%) | 163 (56%) | 81 (78%) |
|
| |||||
| Dropout | |||||
| No dropout | 157 (84%) | 509 (84%) | 467 (83%) | 232 (79%) | 83 (79%) |
| After 60M | 7 (4%) | 24 (4%) | 18 (3%) | 7 (2%) | 2 (2%) |
| After 48M | 10 (5%) | 37 (6%) | 37 (7%) | 25 (9%) | 5 (5%) |
| After 36M | 4 (2%) | 8 (1%) | 14 (2%) | 6 (2%) | 5 (5%) |
| After 24M | 9 (5%) | 11 (2%) | 13 (2%) | 10 (3%) | 5 (5%) |
| After 12M | 0 (0%) | 17 (3%) | 14 (2%) | 12 (4%) | 5 (5%) |
|
| |||||
| TKR | 1 (1%) | 29 (5%) | 78 (14%) | 62 (21%) | 21 (20%) |
| Prior to 12 Month | 0 | 5 | 4 | 4 | 2 |
| Prior to 24 Month | 0 | 2 | 9 | 9 | 7 |
| Prior to 36 Month | 0 | 6 | 18 | 12 | 4 |
| Prior to 48 Month | 0 | 3 | 16 | 13 | 3 |
| Prior to 60 Month | 1 | 7 | 14 | 13 | 3 |
| Prior to 72 Month | 0 | 6 | 17 | 11 | 2 |
|
| |||||
| Medications* | |||||
| Always | 3 (2%) | 19 (3%) | 74 (14%) | 50 (18%) | 37 (38%) |
| Intermittent | 55 (29%) | 263 (45%) | 301 (55%) | 187 (67%) | 54 (55%) |
| None | 129 (69%) | 302 (52%) | 171 (31%) | 43 (15%) | 7 (7%) |
Any medication for pain, aching or stiffness more than half the days of a month, past 12 months, over course of follow-up
Incorporating the covariates into the estimation of the trajectories had minimal impact on the shape of the trajectories. Results of multivariable multinomial regression analysis showed that compared with subjects in the reference trajectory of no pain, subjects in severe pain were more likely to have a higher KL grade, be obese, be younger, be female, have depression, be less educated, have medical comorbidities, and be non-white (Table 3, Supplementary Figure 2). Compared with subjects in no pain, subjects in high moderate pain were more likely to have a higher KL grade, be younger, be female, have depression, be less educated, and have comorbidities. Compared with subjects in no pain, subjects in low moderate pain were more likely to have a higher KL grade, be female, and have depression. We did not identify any covariates under consideration that were significantly associated with belonging in mild pain vs. no pain.
Table 3.
Multivariable Regression Analysis. Odds ratio (95% CI) for belonging in each trajectory relative to reference trajectory (Trajectory 1).
| Mild (n=606) | Low moderate (n=563) | High moderate (n=292) | Severe (n=105) | |
|---|---|---|---|---|
| KL | ||||
| 2 | ref | ref | ref | ref |
| 3 | 1.0 (0.6 – 1.6) | 2.6 (1.7 – 4.1) | 2.6 (1.6 – 4.3) | 4.3 (2.1 – 8.6) |
| 4 | 3.6 (0.9 – 15.0) | 21.9 (6.1 – 79.1) | 25.5 (6.8 – 95.8) | 80.6 (18.7 – 346.8) |
|
| ||||
| Obesity | ||||
| Non-obese | ref | ref | ref | ref |
| Obese | 0.9 (0.6 – 1.4) | 1.3 (0.9 – 1.9) | 1.5 (0.9 – 2.3) | 2.3 (1.2 – 4.4) |
|
| ||||
| Age | 0.98 (0.96 – 1.00) | 0.98 (0.96 – 1.00) | 0.97 (0.95 – 1.00) | 0.92 (0.89 – 0.96) |
|
| ||||
| Sex | ||||
| Male | ref | ref | ref | ref |
| Female | 1.3 (0.8 – 2.0) | 1.6 (1.0 – 2.3) | 2.0 (1.3 – 3.2) | 3.0 (1.5 – 6.2) |
|
| ||||
| Depression | ||||
| CES-D < 16 | ref | ref | ref | ref |
| CES-D >=16 | 1.1 (0.3 – 3.8) | 3.1 (1.2 – 7.9) | 3.9 (1.5 – 10.5) | 8.8 (3.1 – 25.2) |
|
| ||||
| Education | ||||
| College Education | ref | ref | ref | ref |
| Less than college | 0.9 (0.6 – 1.4) | 1.2 (0.8 – 1.8) | 1.7 (1.1 – 2.6) | 5.1 (2.3 – 11.2) |
|
| ||||
| Comorbidities | ||||
| Zero | ref | ref | ref | ref |
| 1+ | 1.0 (0.6 – 1.7) | 1.3 (0.8 – 2.2) | 2.1 (1.3 – 3.5) | 2.0 (1.0 – 3.9) |
|
| ||||
| Race | ||||
| White | ref | ref | ref | ref |
| Non-White | 0.6 (0.4 – 1.1) | 0.9 (0.6 – 1.5) | 1.5 (0.9 – 2.6) | 3.3 (1.7 – 6.6) |
Total Knee Replacement
Participants with early TKRs tended to start with higher WOMAC pain scores at baseline, and on average participants with TKRs tended to experience a worsening in pain in the visits leading up to their surgeries (Figure 2). Undergoing TKR was significantly associated with being in a higher trajectory; including information about TKR in trajectory estimation did not change the size or shape of any trajectory.
Figure 2. Mean WOMAC Pain by TKR Status and Visit.
Mean WOMAC pain is along the Y-axis and follow-up month is along the X-axis. Subjects are split into 7 groups based on when TKR was reported and each line represents one of these groups. The lines show mean WOMAC pain over time. The black line with a circle symbol shows the group that did not have TKR. The green line with the square symbol shows the group that had TKR prior to month 72. The navy line with the diamond symbol shows the group that had TKR prior to month 60. The aqua line with the octagon symbol shows the group that had TKR prior to month 48. The maroon line with the triangle symbol shows the group that had TKR prior to month 36. The green line with the triangle symbol shows the group that had TKR prior to month 24. The pink square symbol shows the group that had TKR prior to month 12.
Secondary Analysis
Analyzing the data by baseline KL did not reveal additional trajectories; the shape of the trajectories was largely the same across KL groups. Participants with higher KL tended to have more pain. We identified only 4 trajectories for the baseline KL 4 group and did not identify a “no pain” trajectory (Figure 3). Fourteen percent of participants with KL 2 at baseline and 23% of participants with KL 3 at baseline progressed to a higher KL by year 4; these trends in progression were not met with corresponding increases in pain for either baseline KL group.
Figure 3. Group-Based Trajectories by Baseline KL Group.
Each of the 3 plots displays a different Baseline KL group. For each graph, WOMAC pain is along the Y-axis and follow-up month is along the X-axis. Each line represents one distinct trajectory. The labels next to each trajectory indicate the percent of the population in that trajectory.
The results of the WOMAC function analysis were similar to the WOMAC pain analysis. Six distinct trajectories were identified, with none of the trajectories showing substantial worsening or improvement over time.
Including the 381 subjects with baseline radiographic OA but without baseline pain did not change the number or shape of the trajectories (Supplementary Figure 3). However, the proportion of subjects in the lower pain trajectories increased. Mean baseline WOMAC pain for the “no pain” group decreased from 15 to 4. Of the 381 subjects, 61% were in the “no pain” trajectory, 34% “mild pain” trajectory, and 4% “low moderate” trajectory.
Discussion
We used group-based trajectory modeling to identify distinct longitudinal pain trajectories in a cohort of persons with radiographic knee OA. We identified 5 distinct trajectories, but none of these demonstrated persistent worsening or improvement over time, with the exception of improvement after a baseline flare. Trajectories were defined with respect to pain measured at baseline. Age, sex, BMI, comorbidities, KL grade, education, race, and depression were all associated with greater pain. Secondary analyses investigating pain trajectories by baseline KL and function did not identify trajectories with persistent worsening or improvement over time.
Most previous work in this area has focused on pain at the cohort level. On average, subjects experience modest worsening over time, although individual results are varied; some subjects experience improvement in symptoms while others experience severe worsening10,11. One study suggested a pain/mental health cycle, where pain leads to depression and fatigue which in turn leads to worsening of pain and function25. Other studies have suggested links between radiographic characteristics and disease progression, with the presence of bone marrow lesions, synovitis, and inflammation increasing the risk of symptomatic progression26,27.
Two previous studies have used a group-based modeling approach to identify outcome trajectories in OA. Leffondré et al. identified four patterns in a cohort of 835 subjects with knee and hip OA over 5 years of follow-up: (1) regularly increasing (18% of cohort), (2) regularly decreasing (24%), (3) stable (40%), and (4) unstable with fluctuations (18%)28. Verkleij et al. examined a cohort of 222 subjects with hip OA with follow-up every 3 months over a 2 year period. Five distinct trajectories were identified: (1) stable mild pain (31%), (2) stable moderate to severe pain (14%), (3) stable severe pain (14%), (4) mild to moderate progressive pain (22%), and (5) mild to severe progressive pain (19%)29. While these studies showed that some subjects remained stable over time and others fluctuated or progressed, our study was unable to identify any trajectories characterized by increasing or decreasing pain over the course of follow-up. Our study evaluated only persons with knee OA, while Leffondré et al. included both knees and hips and Verkleij et al. included only hips. In addition, both studies included subjects with more severe baseline symptoms; the Leffondré et al. cohort had an average WOMAC total score of 39 out of 100 at baseline, while the Verkleij et al. cohort reported an average baseline WOMAC pain of 34 and 9% underwent hip replacement over 2 years of follow-up. In our OAI cohort, average baseline WOMAC total and pain scores were 26 and 27, respectively. Only 2.4% of the cohort underwent TKR by the 2 year visit.
These results have important implications for the way we understand OA as a disease of structural progression and potentially non-progressive symptoms. While pain fluctuated over time, our observation that subject pain did not tend to change over 6 years of follow-up is more consistent with the characterization of OA as a disease of chronic rather than progressive symptoms. Because one might expect concurrent symptomatic and structural progression, our observations of nonparallel radiographic and symptomatic progression have important implications for how we conceptualize the relationship between OA structure and symptomatology. Indeed, while our secondary analysis examined trajectories separately for each KL group, we did not identify any trajectories with persistent worsening, despite the fact that 14% of participants with KL 2 at baseline and 23% of subjects with KL 3 at baseline progressed to a higher KL grade within 4 years.
The relationship between functional decline and chronic rather than episodic pain has been previously explored in the context of other chronic pain conditions, particularly in literature evaluating chronic pain and psychological factors such as pain-related fear30–33. The model describes how a person who feels anxiety in response to pain learns to avoid pain-causing behaviors, leading to the avoidance of physical activity and the subsequent worsening of the underlying disease through the disuse and atrophying of muscles. While this model has been frequently discussed in the context of chronic lower back pain30,32,34, recent studies have provided insight into the relationship between pain-related fear, disability, and knee OA33,35,36. The model reminds us of a basic underlying challenge in longitudinal pain reporting where subjects may inadvertently mask pain progression through the modification of the activities they perform in their daily lives or through the consumption of gradually increasing doses of medications over time. What may appear as a stable pattern of pain may actually reflect increasing pain-related activity limitation and greater reliance upon analgesics in order to maintain a bearable level of pain.
Seventeen percent of the cohort dropped out of the study while 11% underwent TKR over the course of follow-up. The group-based trajectory model assumes that missing data are missing at random. Conditional on the observed outcomes and covariates, missingness does not depend on unobserved outcomes; in other words, attrition and trajectory group assignment are independent37. The basic trajectory model does not accommodate data that are missing not at random, meaning the missing data mechanism depends on the unobserved outcome. In this study, missing not at random would imply that participants’ decisions to drop out or undergo TKR depends on their unobserved pain, even after taking into account the longitudinal pain information and covariates observed until the time of dropout. While we attempted to explore the association between TKR and trajectory by including TKR as a risk factor in our model, further work is needed to explore the impact of missing not at random on pain trajectories. Naturally, TKR is more likely to occur among those in trajectories characterized by greater pain levels.
This study has several important limitations. While group-based trajectory modeling is an established method for analyzing longitudinal data, decisions regarding the number and shape of trajectories are somewhat subjective12. Though we identified 5 trajectories, there was variation in pain in each trajectory. These trajectories do not imply that no individuals worsen or improve over the course of follow-up; rather, we found no pattern of improvement or worsening on average within the trajectory-specific cohorts. The group-based trajectory method, using SAS PROC TRAJ, assumes that any individual deviation from the group-specific trajectory is due to random error, that is, there is no subject-specific random effect. We may wish to investigate methods that allow individual subjects to vary around group-specific trajectories. Secondly, the OAI cohort has been shown to have a higher self-reported health status compared to a nationally representative survey of US persons with OA38. The OAI exclusion criteria, including severe joint space narrowing in both knees (OARSI joint space narrowing grade 3 or bone-on-bone), inability to undergo MRI due to size limits (285 pounds for men and 250 pounds for women), and comorbid conditions that could interfere with participation, may have created a relatively healthy OA cohort. This may help to explain our finding that older age was associated with being in a less severe pain trajectory; perhaps older patients in more severe pain were excluded due to more severe radiographic disease or other comorbid conditions. Further, the younger subjects are likely more physical active and therefore more likely to experience use-related pain. Our results may not be generalizable to the entire population of persons with OA. In fact, our OAI cohort had less severe OA symptoms at baseline than two previous studies investigating symptom trajectories OA cohorts28,29. Thirdly, as this was an observational study, subjects were allowed to continue with their usual treatment for OA symptoms, including pain medications, injections, and physical therapy.
Our study identified 5 distinct pain trajectories in persons with knee OA: no pain, mild pain, low moderate pain, high moderate pain, and severe pain. None of the trajectories demonstrated marked improvement or worsening over time, with the exception of a slight improvement after baseline for all trajectories. Further work is needed to validate these findings in an independent cohort. These findings suggest that some people may experience OA as a chronic rather than progressive disease.
Supplementary Material
Acknowledgments
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 was prepared using an OAI public use data set and does not necessarily reflect the opinions or views of the OAI investigators, the NIH, or the private funding partners.
ROLE OF THE FUNDING SOURCE
Supported by: NIH/NIAMS T32AR055885, R01 AR 064320, K24 AR 057827, and P60 AR 047782.
Footnotes
CONFLICT OF INTEREST
The authors do not have any conflict of interest with respect to the context of this paper.
AUTHOR CONTRIBUTIONS
Conception and design: Collins, Losina
Analysis and interpretation of the data: Collins, Katz, Losina
Drafting of the article: Collins, Dervan
Critical revision of the article for important intellectual content: Collins, Katz, Losina, Dervan
Final approval of the article: Collins, Katz, Dervan, Losina
Provision of study materials or patients: OAI
Statistical expertise: Collins, Losina
Obtaining of funding: Losina
Collection and assembly of data: Collins, OAI
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
References
- 1.Felson DT, Lawrence RC, Dieppe PA, Hirsch R, Helmick CG, Jordan JM, et al. Osteoarthritis: new insights. Part 1: the disease and its risk factors. Annals of Internal Medicine. 2000;133:635–646. doi: 10.7326/0003-4819-133-8-200010170-00016. [DOI] [PubMed] [Google Scholar]
- 2.Dunlop DD, Manheim LM, Song J, Chang RW. Arthritis prevalence and activity limitations in older adults. Arthritis & Rheumatism. 2001;44:212–221. doi: 10.1002/1529-0131(200101)44:1<212::AID-ANR28>3.0.CO;2-Q. [DOI] [PubMed] [Google Scholar]
- 3.Salaffi F, Carotti M, Stancati A, Grassi W. Health-related quality of life in older adults with symptomatic hip and knee osteoarthritis: a comparison with matched healthy controls. Aging clinical and experimental research. 2005;17:255. doi: 10.1007/BF03324607. [DOI] [PubMed] [Google Scholar]
- 4.Gupta S, Hawker G, Laporte A, Croxford R, Coyte P. The economic burden of disabling hip and knee osteoarthritis (OA) from the perspective of individuals living with this condition. Rheumatology. 2005;44:1531–1537. doi: 10.1093/rheumatology/kei049. [DOI] [PubMed] [Google Scholar]
- 5.MacLean CH, Knight K, Paulus H, Brook RH, Shekelle PG. Costs attributable to osteoarthritis. Journal of Rheumatology. 1998;25:2213–2218. [PubMed] [Google Scholar]
- 6.Hadler NM. Knee pain is the malady—not osteoarthritis. Annals of Internal Medicine. 1992;116:598–599. doi: 10.7326/0003-4819-116-7-598. [DOI] [PubMed] [Google Scholar]
- 7.Dieppe PA, Lohmander LS. Pathogenesis and management of pain in osteoarthritis. The Lancet. 2005;365:965–973. doi: 10.1016/S0140-6736(05)71086-2. [DOI] [PubMed] [Google Scholar]
- 8.van Dijk GM, Dekker J, Veenhof C, van den Ende CH. Course of functional status and pain in osteoarthritis of the hip or knee: a systematic review of the literature. Arthritis Rheum. 2006;55:779–785. doi: 10.1002/art.22244. [DOI] [PubMed] [Google Scholar]
- 9.Felson D, Niu J, Sack B, Aliabadi P, McCullough C, Nevitt MC. Progression of osteoarthritis as a state of inertia. Ann Rheum Dis. 2012 doi: 10.1136/annrheumdis-2012-201575. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Peters TJ, Sanders C, Dieppe P, Donovan J. Factors associated with change in pain and disability over time: a community-based prospective observational study of hip and knee osteoarthritis. Br J Gen Pract. 2005;55:205–211. [PMC free article] [PubMed] [Google Scholar]
- 11.Dieppe P, Cushnaghan J, Tucker M, Browning S, Shepstone L. The Bristol ‘OA500 study’: progression and impact of the disease after 8 years. Osteoarthritis and Cartilage. 2000;8:63–68. doi: 10.1053/joca.1999.0272. [DOI] [PubMed] [Google Scholar]
- 12.Nagin D. Group-based modeling of development. Harvard University Press; 2005. [Google Scholar]
- 13.Kellgren J, Lawrence J. Radiological assessment of osteo-arthrosis. Ann Rheum Dis. 1957;16:494–502. doi: 10.1136/ard.16.4.494. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Bellamy N, Buchanan WW, Goldsmith CH, Campbell J, Stitt L. Validation study of WOMAC: a health status instrument for measuring clinically important patient relevant outcomes to antirheumatic drug therapy in patients with osteoarthritis of the hip or knee. J Rheumatol. 1988;15:1833. [PubMed] [Google Scholar]
- 15.Angst F, Aeschlimann A, Stucki G. Smallest detectable and minimal clinically important differences of rehabilitation intervention with their implications for required sample sizes using WOMAC and SF-36 quality of life measurement instruments in patients with osteoarthritis of the lower extremities. Arthritis Rheum. 2001;45:384–391. doi: 10.1002/1529-0131(200108)45:4<384::AID-ART352>3.0.CO;2-0. [DOI] [PubMed] [Google Scholar]
- 16.Tubach F, Ravaud P, Baron G, Falissard B, Logeart I, Bellamy N, et al. Evaluation of clinically relevant changes in patient reported outcomes in knee and hip osteoarthritis: the minimal clinically important improvement. Ann Rheum Dis. 2005;64:29–33. doi: 10.1136/ard.2004.022905. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Zelman DC, Hoffman DL, Seifeldin R, Dukes EM. 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;106:35–42. doi: 10.1016/s0304-3959(03)00274-4. [DOI] [PubMed] [Google Scholar]
- 18.Kapstad H, Hanestad BR, Langeland N, Rustoen T, Stavem K. Cutpoints for mild, moderate and severe pain in patients with osteoarthritis of the hip or knee ready for joint replacement surgery. BMC Musculoskelet Disord. 2008;9:55. doi: 10.1186/1471-2474-9-55. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Nevitt MC, Felson D, Lester G. The Osteoarthritis Initiative: Protocol for the Cohort Study. University of California; San Francisco: 2006. [Google Scholar]
- 20.Katz JN, Chang LC, Sangha O, Fossel AH, Bates DW. Can comorbidity be measured by questionnaire rather than medical record review? Medical care. 1996;34:73–84. doi: 10.1097/00005650-199601000-00006. [DOI] [PubMed] [Google Scholar]
- 21.Radloff LS. The CES-D scale A self-report depression scale for research in the general population. Applied psychological measurement. 1977;1:385–401. [Google Scholar]
- 22.Dodge HH, Shen C, Ganguli M. Application of the Pattern-Mixture Latent Trajectory Model in an Epidemiological Study with Non-Ignorable Missingness. J Data Sci. 2008;6:247–259. [PMC free article] [PubMed] [Google Scholar]
- 23.Jones BL, Nagin DS. Advances in group-based trajectory modeling and an SAS procedure for estimating them. Sociological Methods & Research. 2007;35:542–571. [Google Scholar]
- 24.Jones BL, Nagin DS, Roeder K. A SAS procedure based on mixture models for estimating developmental trajectories. Sociological Methods & Research. 2001;29:374–393. [Google Scholar]
- 25.Hawker GA, Gignac MA, Badley E, Davis AM, French MR, Li Y, et al. A longitudinal study to explain the pain-depression link in older adults with osteoarthritis. Arthritis Care Res (Hoboken) 2011;63:1382–1390. doi: 10.1002/acr.20298. [DOI] [PubMed] [Google Scholar]
- 26.Stannus OP, Jones G, Blizzard L, Cicuttini FM, Ding C. Associations between serum levels of inflammatory markers and change in knee pain over 5 years in older adults: a prospective cohort study. Ann Rheum Dis. 2013;72:535–540. doi: 10.1136/annrheumdis-2011-201047. [DOI] [PubMed] [Google Scholar]
- 27.Zhang Y, Nevitt M, Niu J, Lewis C, Torner J, Guermazi A, et al. Fluctuation of knee pain and changes in bone marrow lesions, effusions, and synovitis on magnetic resonance imaging. Arthritis Rheum. 2011;63:691–699. doi: 10.1002/art.30148. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Leffondre K, Abrahamowicz M, Regeasse A, Hawker GA, Badley EM, McCusker J, et al. Statistical measures were proposed for identifying longitudinal patterns of change in quantitative health indicators. J Clin Epidemiol. 2004;57:1049–1062. doi: 10.1016/j.jclinepi.2004.02.012. [DOI] [PubMed] [Google Scholar]
- 29.Verkleij SP, Hoekstra T, Rozendaal RM, Waarsing JH, Koes BW, Luijsterburg PA, et al. Defining discriminative pain trajectories in hip osteoarthritis over a 2-year time period. Ann Rheum Dis. 2012;71:1517–1523. doi: 10.1136/annrheumdis-2011-200687. [DOI] [PubMed] [Google Scholar]
- 30.Lethem J, Slade PD, Troup JD, Bentley G. Outline of a Fear-Avoidance Model of exaggerated pain perception--I. Behav Res Ther. 1983;21:401–408. doi: 10.1016/0005-7967(83)90009-8. [DOI] [PubMed] [Google Scholar]
- 31.Vlaeyen JW, Linton SJ. Fear-avoidance and its consequences in chronic musculoskeletal pain: a state of the art. Pain. 2000;85:317–332. doi: 10.1016/S0304-3959(99)00242-0. [DOI] [PubMed] [Google Scholar]
- 32.Leeuw M, Goossens ME, Linton SJ, Crombez G, Boersma K, Vlaeyen JW. The fear-avoidance model of musculoskeletal pain: current state of scientific evidence. J Behav Med. 2007;30:77–94. doi: 10.1007/s10865-006-9085-0. [DOI] [PubMed] [Google Scholar]
- 33.Heuts PH, Vlaeyen JW, Roelofs J, de Bie RA, Aretz K, van Weel C, et al. Pain-related fear and daily functioning in patients with osteoarthritis. Pain. 2004;110:228–235. doi: 10.1016/j.pain.2004.03.035. [DOI] [PubMed] [Google Scholar]
- 34.Vlaeyen JW, Linton SJ. Fear-avoidance model of chronic musculoskeletal pain: 12 years on. Pain. 2012;153:1144–1147. doi: 10.1016/j.pain.2011.12.009. [DOI] [PubMed] [Google Scholar]
- 35.Wright LJ, Zautra AJ, Going S. Adaptation to early knee osteoarthritis: the role of risk, resilience, and disease severity on pain and physical functioning. Ann Behav Med. 2008;36:70–80. doi: 10.1007/s12160-008-9048-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Scopaz KA, Piva SR, Wisniewski S, Fitzgerald GK. Relationships of fear, anxiety, and depression with physical function in patients with knee osteoarthritis. Arch Phys Med Rehabil. 2009;90:1866–1873. doi: 10.1016/j.apmr.2009.06.012. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Haviland AM, Jones BL, Nagin DS. Group-based Trajectory Modeling Extended to Account for Nonrandom Participant Attrition. Sociological Methods & Research. 2011;40:367–390. [Google Scholar]
- 38.Reichmann WM, Katz JN, Losina E. Differences in self-reported health in the Osteoarthritis Initiative (OAI) and Third National Health and Nutrition Examination Survey (NHANES-III) PLoS One. 2011;6:e17345. doi: 10.1371/journal.pone.0017345. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.