Quantifying Sustained Pain Worsening in Knee Osteoarthritis

Abstract

Objective:

Recent work suggests that many persons with knee osteoarthritis (OA) experience stable symptoms over time. Whether patients experience periods of symptom exacerbation or flare which interrupt this stable course, and how long such periods last, has received little study. Our objective is to describe the frequency and duration of episodes of pain worsening in persons with knee OA.

Methods:

We selected participants from the Osteoarthritis Initiative with radiographic, symptomatic knee OA. We defined a clinically relevant increase in knee pain as an increase in WOMAC pain of ≥9 points. We defined sustained worsening as maintaining at least 80% of the initial increase. We used Poisson regression to estimate the incidence rate (IR) of episodes of pain worsening.

Results:

1093 participants were included in the analysis. Eighty-eight percent had ≥1 increase in WOMAC pain ≥ 9 points (IR: 26.3 per 100 person years (95% CI: 25.2, 27.4)). Forty-eight percent had ≥1 episode of sustained worsening (IR: 9.7 per 100 person-years (95% CI: 8.9, 10.5)). Elevated pain was maintained an average of 2.4 years after the initial increase.

Conclusion:

Most participants with knee OA reported at least one clinically relevant increase in WOMAC pain, but fewer than half experienced an episode of sustained pain worsening. These individual-level data portray a more nuanced and fluctuating course of OA pain than suggested by trajectory studies. These data could be useful in shared decision-making regarding prognosis and treatment choices in persons affected by symptomatic knee OA.

Introduction:

Knee osteoarthritis (OA) affects 14 million individuals in the United States and over 300 million adults worldwide.^{1, 2} The disease is characterized by pain, aching, and stiffness and is the leading source of disability in people older than 55 years of age. Nearly half of adults with doctor diagnosed arthritis in the US report arthritis-attributable activity limitations.³ Knee pain is the primary reason that persons with knee OA seek medical care.⁴

While traditionally considered a disease with slow, inevitable progression, recent work suggests that many patients with knee OA report relatively stable pain and function over 5+ years of follow-up.^5–7 A recent systematic review of pain trajectories in knee OA found that approximately 85% of knee OA patients are classified in a stable pain trajectory over the medium to long term (5 – 8 years).⁸ However, these stable trajectories may include patients with fluctuating symptoms such as patients who experience a pattern of flare and remission. Parry et al. reported that almost one-third of knee osteoarthritis patients experienced substantial pain variability over 72 months.⁹

Recently, experts have proposed reconceptualizing OA as an ‘acute-on-chronic’ condition similar to chronic obstructive pulmonary disease or gout.¹⁰ In such conditions, patients experience periods of symptom exacerbation or flare which interrupt the underlying chronic course of the disease. Understanding within-patient variability in the course of symptoms may shed light on patients’ pain experiences. In particular, understanding whether clinically important symptom exacerbations tend to resolve or result in episodes of sustained worsening may help to inform patient expectations and treatment strategies in persons affected by knee OA. Our objectives were to describe the frequency and duration of pain worsening episodes in persons with knee OA and to evaluate whether baseline disease severity is associated with sustained pain worsening.

Methods:

Sample

We used data from the Osteoarthritis Initiative (OAI), a multi-center, longitudinal observational study of knee OA. Men and women ages 45–79 with or at high risk of developing knee OA were enrolled at four centers between 2004 and 2006. Subjects completed questionnaires annually for 9 years. The data and additional study details are publicly available.¹¹

We included subjects with Kellgren-Lawrence (KL) grade ≥2 and frequent knee symptoms.¹² Frequent knee symptoms were defined as “pain, aching or stiffness in or around the knee on most days” for at least one month during the past 12 months.”¹³ KL grade was ascertained from the central reading of a standardized fixed-flexion radiograph. Frequent knee symptoms were defined as pain, aching, or stiffness in or around the knee on most days for at least one month during the 12 months preceding study enrollment. Because we focused on estimating episodes of sustained pain worsening, we required that participants had enough follow-up time to observe such episodes; therefore, we included participants with at least two years of follow-up.

Participants unable to experience pain worsening ≥ 9 points because of extremely high (i.e., ≥91) WOMAC pain scores were also excluded. We included one knee per participant; in the case that both knees met the inclusion criteria, we selected the knee with worse WOMAC pain and randomly selected a knee if the WOMAC pain was the same in both knees.

Outcome

Our primary outcome was pain worsening based on the pain subscale from the WOMAC index, scaled from 0–100 with higher scores indicating more severe pain.¹⁴ The WOMAC pain index asks participants to rate their pain “over the last 7 days” on a 5-point Likert scale from “none” to “extreme” pain for the following activities: walking, stairclimbing, sitting, lying down, and standing. We considered a 9-point increase over 1 year a clinically meaningful increase in WOMAC pain, and we noted this increase as pain worsening for the purposes of this analysis.¹⁵ This threshold is based on the literature for a minimum clinically important difference for pain worsening.^15–17 Whether this initial ≥9 point increase is the start of an episode of sustained pain worsening is described below.

Episodes of Sustained Pain Worsening

We defined an episode of sustained pain worsening as a time interval over which a clinically meaningful increase in WOMAC pain was maintained for at least one year. In our primary analysis we considered sustained pain worsening as maintaining at least 80% of the initial increase in pain. For example, a subject with a pain score of 20 at baseline increasing to a score of 35 at year 1 would have to maintain at least 80% of this increase (e.g., 0.80 × 15 = 12 points) at year 2 (e.g., maintain a WOMAC score of 20 + 12 = 32) to have a sustained episode of pain worsening. An episode of sustained pain worsening could start from any follow-up assessment point. For each pain worsening episode, we counted the number of years that this elevated pain score was maintained. Participants could have more than one episode of pain worsening, but could not begin a new episode of sustained pain worsening while already in an episode. As a sensitivity analysis, we defined an episode of sustained pain worsening as maintaining a ≥9 point increase from starting pain. For example, a subject with a pain score of 20 at baseline increasing to a score of 35 at year 1 must maintain a WOMAC score of ≥29 in order to have a sustained episode of pain worsening.

Examples of pain worsening for two participants are shown in Figure 1. Participant 1 increases from a WOMAC pain score of 20 to 30 from year 1 to year 2. This increase is maintained at year 3 with a WOMAC pain score of 35, signaling an episode of sustained worsening. The episode is maintained until year 7, when pain returns to a score of 25. Participant 2 experiences two increases in WOMAC pain ≥ 9 points, from years 1 to 2 and 7 to 8. These increases in WOMAC pain are not maintained the following year; thus, this participant experienced no episodes of sustained pain worsening.

Figure 1.

Examples of Pain Worsening

Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) pain over time for two study participants are shown in the solid dark gray line with squares (participant 1) and the dashed light gray line with circles (participant 2). WOMAC pain is along the y-axis and time point is along the x-axis. The symbols represent the reported WOMAC pain at each timepoint.

Demographic and Clinical Characteristics

Additional demographic and clinical information included baseline age, sex, body mass index (BMI), comorbidities (Charlson Comorbidity Index), race, depression (Center for Epidemiologic Studies Depression Scale (CES-D) Score),¹⁸ KL grade, history of knee injury, and prior knee surgery. History of knee injury was assessed with the question, “Have you ever injured your knee badly enough to limit ability to walk for at least two days?” History of knee surgery was assessed with the question, “Did you ever have surgery or arthroscopy on your knee?” Baseline WOMAC pain was grouped into the following categories: ≤15, 16–40, >40.

Statistical Analysis

We used a piecewise linear mixed model to describe the overall trend in WOMAC pain over time. We used an unstructured covariance matrix to account for repeated measurements over time within person and included a random effect for study site. We included a knot at year 1 to model the change in WOMAC pain from baseline to year 1 and from year 1 to year 9 separately to capture regression to the mean. We chose a knot at year 1 based on visual inspection of WOMAC pain over time (Figure A1). We examined residuals to confirm the appropriateness of a linear model for WOMAC pain. We used repeated measures Poisson regression to compute the overall incidence rate (IR) of pain worsening per 100 person-years with 95% confidence intervals (CIs). We censored participants at the time of drop-out, lost-to-follow-up, TKR, or end of study (year 9). In the analyses of any pain worsening, a participant was eligible to experience the outcome of interest at any visit after baseline; person-time ranged from 2 (inclusion criterion for this analysis) to 9 (final OAI visit) years. In the analyses of sustained pain worsening, participants were not eligible to begin a new episode of sustained pain worsening while already in an episode. Participants could also not begin a new episode of sustained pain worsening in the final year of follow-up, as at least two years of follow-up are necessary to determine whether an increase in pain is sustained for more than one year. Person-time therefore excluded years in which a participant was in an episode of sustained pain worsening and participants’ final year of follow-up. We evaluated the association between baseline KL and pain group in separate unadjusted Poisson regression models and in models including both factors. We calculated the number of episodes of sustained pain worsening that were maintained to the end of follow-up. To estimate the length of episodes we subset on episodes that began with at least 3 years of follow-up remaining.

We imputed missing WOMAC pain data using multiple imputation (MI) based on the multivariate normal distribution.¹⁹ We imputed data under the missing at random assumption, assuming that missing pain data were associated with observed pain data. 20 The imputation model included all observed pain data. We did not impute missing data after a subject dropped out of the study, was lost to follow-up, or underwent TKR. Approximately 4% of observations were imputed. The number of imputed observations is provided in Appendix Table 2. We generated 10 imputed datasets so that the number of imputations was greater than the percent of missing data.²¹ We used the MIANALYZE procedure in SAS 9.4 to combine the results across the imputations using Rubin’s rules.^{20, 22} All analyses of worsening (any worsening, sustained worsening) were conducted on the imputed dataset.

All analyses were conducted using SAS 9.4 (SAS Institute, Cary NC).

Results:

Cohort characteristics

There were 1171 OAI participants with radiographic, symptomatic OA at baseline; 1094 (93%) of these had follow-up WOMAC pain data through at least year 3. One participant was ineligible to meet our definition of worsening due to high WOMAC pain (≥91 at the first three of four visits) and was excluded. The analytic sample included 1093 participants. Participants excluded from the analysis were on average older, more likely to have KL 4, and had higher baseline WOMAC pain compared to participants included in the analysis (Supplemental Table A1). Approximately half of the excluded participants underwent TKR prior to the two-year visit. The mean age of those included in the analysis was 61.5 (SD 9.1), 57% female, 71% White (Table 1). Forty-nine percent were KL 2, 37% KL 3, and 14% KL 4. Mean baseline WOMAC pain was 30 (SD 20). Participants were followed for a median of 9.0 years and mean of 7.5 years.

Table 1.

Baseline Characteristics for the Analytic Sample. Presented in cells: mean (SD) for continuous variables, n (%) for categorical variables.

Cohort Characteristics
Age	61.5 (9.1)
Gender
Male	476 (43.5%)
Female	617 (56.5%)
BMI	30.2 (4.9)
Number of Comorbidities
0	760 (70.6%)
1	189 (17.6%)
2+	127 (11.8%)
Racial background
Asian	7 (0.6%)
Black or African American	290 (26.5%)
White or Caucasian	773 (70.7%)
Other	23 (2.1%)
Center for Epidemiologic Studies Depression Scale (CES-D) Score KLG	7.8 (7.9)
2	536 (49.0%)
3	401 (36.7%)
4	156 (14.3%)
WOMAC Pain	30.1 (19.6)
Prior knee injury
No	599 (55.1%)
Yes	489 (44.9%)
Prior knee surgery
No	748 (68.5%)
Yes	344 (31.5%)

Pain changes over time

On average, participants’ WOMAC pain decreased by 6.7 points (95% confidence interval (CI): 5.7, 7.6) from baseline to year 1, and then increased by 0.2 points per year (95% CI: 0.1, 0.4) from year 1 to year 9, resulting in an average final pain score of 25. Median WOMAC pain was 20 at each visit from year 1 to year 9. Mean WOMAC pain over time is shown in Supplemental Figure A1.

Eighty-eight percent of participants had at least one increase in WOMAC pain ≥ 9 points. The incidence rate of WOMAC pain increase ≥ 9 points was 26.3 per 100 person years (95% CI: 25.2, 27.4), as shown in Table 2. Among participants with an increase in WOMAC pain ≥ 9 points, the mean increase in WOMAC pain was 19 points (median increase was 15, range 9 – 85).

Table 2.

Incidence Rate of Worsening.

Analysis	Sample size	Total number of episodes of sustained worsening	Mean number of episodes of sustained worsening	Total Follow-up Time (years)	Mean Follow-up Time (years)	Unadjusted Incidence Rate (95% CI) per 100 person-years	Adjusted Incidence Rate (95% CI) per 100 person-years
Any Increase	1093	2149	2.0	8183	7.5	26.3 (25.2, 27.4)	--
Episode of Sustained Worsening	1093	590	0.54	6099	5.6	9.7 (8.9, 10.5)	--
Baseline KL Subgroups
KL 2	536	279	0.52	3280	6.1	8.5 (7.5, 9.6)	8.6 (7.7, 9.7)
KL 3	401	232	0.58	2155	5.4	10.7 (9.4, 12.3)	10.8 (9.5, 12.2)
KL 4	156	79	0.51	664	4.3	11.9 (9.5, 14.9)	11.7 (9.5, 14.3)
Baseline pain Subgroups
≤15	337	156	0.46	2054	6.1	7.5 (6.4, 8.9)	8.2 (6.9, 9.7)
16–40	479	248	0.52	2563	5.3	9.6 (8.5, 11.0)	10.1 (8.9, 11.4)
>40	277	186	0.67	1483	5.4	12.6 (10.8, 14.6)	13.2 (11.6, 15.1)

Incidence of pain worsening episodes

Five hundred and twenty (48%) participants had at least one episode of sustained pain worsening, meaning that at least 80% of the initial clinically meaningful increase of ≥9 points was maintained for at least one year. The incidence rate (IR) of sustained pain worsening was 9.7 per 100 person-years (95% CI: 8.9, 10.5), as shown in Table 2. Episodes began from an average starting pain of 15 (median 10) and the mean initial increase was 18 points (median 15). The mean total increase in pain over the entire episode was 27 points (median 25).

We evaluated the length of episodes for the 419 episodes (71%) that began with at least 3 years of follow-up remaining. Elevated pain was maintained an average of 2.4 years after the initial pain increase (median 2 years, range 1–8 years). Twenty-six percent of episodes were maintained until the end of follow-up (i.e., elevated pain was maintained until the last visit).

Baseline OA severity associated with pain worsening

In models including both baseline KL and baseline pain group the incidence of sustained pain worsening was associated with both factors (Figure 2): the IR per 100 person-years was 8.6 (95% CI: 7.7, 9.7) for participants with baseline KL 2, 10.8 (95% CI: 9.5, 12.2) for KL 3, and 11.7 (95% CI: 9.5, 14.3) for KL 4. The IR per 100 person-years was 8.2 (95% CI: 6.9, 9.7) for those with a baseline WOMAC pain less than or equal to 15, 10.1 (95% CI: 8.9, 11.5) for those with a baseline WOMAC pain between 16 and 40, and 13.2 (95% CI: 11.6, 15.1) for those with a baseline WOMAC pain greater than 40.

Figure 2.

Incidence Rate of Sustained Pain Worsening by Subgroup

Incidence rate (IR) with 95% confidence intervals of Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) pain worsening episodes per 100 person-years as estimated using Poisson regression. Incidence rate (IR) is along the y-axis; subgroups are along the x-axis. Subgroups are separated by baseline pain and Kellgren Lawrence (KL) tibiofemoral osteoarthritis classification scale. KL Grade 2 = minimal osteoarthritis, KL 3 = moderate osteoarthritis, KL 4 = severe osteoarthritis ¹².

Sensitivity analysis

In sensitivity analyses with a less strict criterion for sustained pain worsening (maintaining a ≥9 point increase from starting pain), 672 (61%) of participants had at least one episode of sustained pain worsening. The IR per 100 person-years was 14.1 (95% CI: 13.2, 15.2). The mean total increase in pain over the entire episode was 23 points (median 20). Elevated pain was maintained an average of 2.9 years after the initial pain increase (median 3 years, range 1–8 years). Thirty-seven percent of episodes were maintained until the end of follow-up.

Discussion:

We examined pain over 9 years in a cohort of participants with radiographic, symptomatic knee OA and found that while on average pain was relatively stable over 9 years, most participants reported clinically important symptom fluctuations. While increases in pain were common, only approximately half of pain increases were maintained more than one year, underscoring the variable nature of pain.

Two recent meta-analyses summarized work on pain trajectories in OA. Previtali et al. reviewed 7 studies evaluating knee OA pain trajectories and found that the majority of participants report stable pain over the medium to long term (5–8 years of follow-up).⁸ Interestingly, the probability of belonging to a trajectory with increasing pain (6.7%) was similar to that of belonging to a trajectory with decreasing pain (7.9%).⁸ Wieczorek et al. reviewed 21 studies evaluating pain trajectories in persons with knee or hip OA.²³ The authors found high heterogeneity in terms of patient characteristics (e.g., early OA vs. established OA), outcomes assessed, and follow-up time, but found that trajectories were relatively stable over time and most patients belonged to subgroups with mild or low moderate pain.²³

Work is ongoing to define pain flare in OA; the Outcome Measures in Rheumatology (OMERACT) Flares in Knee and Hip Osteoarthritis Working Group described pain during flare as “a distinct change in pain, that is more severe and lasts longer, that is particularly heightened with physical activity and persists with rest”.²⁴ Much of the work examining pain flare in OA has assessed acute pain flare over the short term; few studies have evaluated how pain flares or symptom exacerbations fit in with the long term pain experience.²⁵ Parry et al. evaluated pain variability over 72 months in participants with knee OA using an 11-point numeric rating scale (NRS).⁹ The authors defined “significant pain variability” as reporting a worst pain intensity of ≥5/10 and at reporting worst pain at least 2 points higher than average pain intensity. Pain was assessed every 18 months. The authors found that 53% of participants met the definition for significant pain variability at one or more timepoints. Average pain intensity remained stable, both for participants meeting the significant pain variability definition (NRS 4.7 at baseline, 4.9 at 72 months) and for those without significant pain variability (NRS 3.9 at baseline, 3.8 at 72 months).⁹

Long term studies of pain in osteoarthritis often include long intervals between assessments, typically 12 to 18 months.^5–7 Combined with the relatively short look-back window of common pain questionnaires (e.g., in the OAI participants were asked to rate their symptoms “over the last 7 days” when completing the WOMAC questionnaire), it is difficult to understand whether assessments of longitudinal pain represent increasing or decreasing pain trajectories or pain flare (defined as “a distinct change in pain, that is more severe and lasts longer, that is particularly heightened with physical activity and persists with rest.”).¹⁰ Studies of pain flare, on the other hand, while collecting data at weekly or even daily intervals, are often short-term, over weeks or months.^26–28 More frequent assessments over longer time intervals would help us to better understand the role of pain variability in long-term trajectory.

Much like Parry et al., we found that while on average pain remained stable over 9 years of follow-up, most participants reported increases in pain exceeding the MCID. In our analysis, we attempted to differentiate between a short-term pain exacerbation that resolves by the following year and increased pain that is maintained and may represent symptomatic disease progression. Approximately half of initial increases in pain were resolved by the following year. Among those increases that were maintained for at least one year, the average time spent in an episode of sustained worsening was 2.4 years; almost 75% of episodes were resolved by the end of follow-up. This work adds to a growing body of literature suggesting that the long-term course of pain in OA is variable.

The results of this analysis should be viewed in light of several limitations. The OAI cohort may not be generalizable to the entire population of patients with OA: it has a higher self-reported health status compared to a nationally representative survey of US patients with OA.²⁹ Additionally, we excluded participants from this analysis who did not have at least two years of follow-up (n=77) or who were ineligible to meet our definition of worsening due to WOMAC pain scores ≥91 (n=1). While this number was small (n=78, 7%), these patients were more likely to have more advanced OA at baseline; half underwent TKR prior to the two-year follow-up. Thus, we may be underestimating the risk of sustained pain worsening in persons with OA. While median follow-up time was 9 years, only approximately 60% of participants completed the year 9 visit. While Poisson regression accounts for variable person-time, we did not explicitly consider a non-random missing data mechanism. Seventeen percent of the analytic cohort underwent TKR over the course of follow-up; it is possible that these patients would have experienced episodes of sustained worsening had they not undergone TKR. We sought to quantify the number of years that each episode of sustained worsening was maintained, but were limited in our ability to do this as many episodes were maintained until the last follow-up visit (i.e., censored). We attempted to address this by estimating episode length only on those episodes that began with at least 3 years of follow-up remaining, but even then, 26% were maintained to the end of follow-up. Thus, our estimates of episode length are likely underestimates. While efforts are underway to better define pain flare in OA, there is no consensus definition of sustained pain worsening.^{24, 30} We considered a definition that required participants to maintain 80% of their initial pain increase. This 80% threshold was somewhat arbitrary, and therefore we evaluated a secondary, less strict, definition in the sensitivity analysis.

The goal of these analyses is to describe the incidence of sustained pain worsening in knee OA, and to describe the association between baseline OA severity and incidence of sustained pain worsening. Felson and colleagues describe structural progression as fitting 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 stable.³¹ In our analyses we found that participants with worse baseline disease severity – both in terms of structure and pain – were more likely to experience episodes of sustained worsening. We ultimately decided not to adjust for potential confounders in order to avoid potential collider bias.³² Research questions involving the interplay between disease severity, risk factors, and outcomes requires careful consideration of the causal model and advanced modeling techniques such as natural effects models to properly model the complex relationships between variables.³³ The OAI was an observational study, subjects were allowed to continue with their usual treatment for OA symptoms, including pain medications, injections, and physical therapy. Future work could consider incorporating time-varying covariates, using an approach such as marginal structural modeling, to investigate what patient factors or treatments are associated with incidence or resolution of symptom exacerbation.

We found that most participants with knee OA reported at least one increase in WOMAC pain exceeding the 9 points, but fewer than half of these participants experienced an episode of sustained pain worsening (i.e., lasting more than 1 year). While the annual risk of an episode of sustained worsening is low, over the approximately 20 years that patients live with knee OA, most patients will likely experience at least one episode of sustained worsening. These individual level data portray a more nuanced and fluctuating course of OA pain than suggested by trajectory studies.

Appendix

Figure A1.

Mean WOMAC Pain by Study Year. Study year is along the X-axis and mean WOMAC pain for the analytic cohort is along the Y-axis. The blue dots represent mean pain at each year and the shaded band represents the 95% confidence interval.

Figure A2.

Mean WOMAC Pain by Study Year AND Sustained Worsening Status. Study year is along the X-axis and mean WOMAC pain is along the Y-axis. Participants are categorized into participants experiencing at least one episode of sustained worsening (dark gray) vs. those that experienced no episodes of sustained worsening (light gray). The line represents mean pain and the shaded band the 95% confidence interval, summarized across 11 imputations.

Table A1:

Comparison of Participants Included vs. Excluded Due to Missing Data. Presented in cells: mean (SD) for continuous variables, n (%) for categorical variables.

Cohort Characteristics	Included in Analytic Cohort
	No (n=78)	Yes (n=1093)
Age	62.6 (8.0)	61.5 (9.1)
Gender
Male	32 (41%)	476 (44%)
Female	46 (59%)	617 (56%)
BMI	31.1 (4.7)	30.2 (4.9)
Number of Comorbidities
0	54 (69%)	760 (71%)
1	12 (15%)	189 (18%)
2+	12 (15%)	127 (12%)
Missing	0	17
Center for Epidemiologic Studies Depression Scale (CES-D) Score KLG	7.4 (8.1)	7.8 (7.9)
2	24 (31%)	536 (49%)
3	27 (35%)	401 (37%)
4	27 (35%)	156 (14%)
WOMAC Pain	38.9 (21.4)	30.1 (19.6)
Prior knee injury
No	35 (45%)	599 (55%)
Yes	42 (55%)	489 (45%)
Prior knee surgery
No	42 (54%)	748 (68%)
Yes	36 (46%)	344 (32%)
TKR between baseline and year 2
No	40 (51%)	1093 (100%)
Yes	38 (49%)	0 (0%)
TKR between baseline and year 9
No	34 (44%)	906 (83%)
Yes	44 (56%)	187 (17%)

Table A2:

Summary of Missing and Imputed Data by Timepoint. Intermittent missing data imputed using multiple imputation (MI) based on multivariate normal distribution. Data were not imputed after dropout/lost-to-follow-up or TKR. The number of each data type for WOMAC pain (observed WOMAC pain score; imputed using MI; missing due to drop-out/lost-to-follow-up, not imputed; missing due to TKR, not imputed) are shown for each year.

Year	Observed	Imputed	Missing: Dropout/Lost-to-Follow-Up	Missing: TKR
0	1093 (100%)	0 (0%)	0 (0%)	0 (0%)
1	1054 (96.4%)	39 (3.6%)	0 (0%)	0 (0%)
2	1038 (95%)	55 (5%)	0 (0%)	0 (0%)
3	976 (89.3%)	53 (4.8%)	32 (2.9%)	32 (2.9%)
4	950 (86.9%)	28 (2.6%)	55 (5%)	60 (5.5%)
5	851 (77.9%)	52 (4.8%)	96 (8.8%)	94 (8.6%)
6	797 (72.9%)	63 (5.8%)	114 (10.4%)	119 (10.9%)
7	770 (70.4%)	53 (4.8%)	128 (11.7%)	142 (13%)
8	767 (70.2%)	5 (0.5%)	154 (14.1%)	167 (15.3%)
9	632 (57.8%)	0 (0%)	274 (25.1%)	187 (17.1%)
Total	8928 (81.7%)	348 (3.2%)	853 (7.8%)	801 (7.3%)