To determine the trajectories of preoperative worsening and postoperative recovery for both the index knee and the contralateral knee of patients undergoing knee replacement surgery.
To determine the trajectories of preoperative worsening and postoperative recovery for both the index knee and the contralateral knee of patients undergoing knee replacement surgery.
Of the 4,796 subjects in the Osteoarthritis Initiative cohort study database, we examined 5-year data from 177 patients who underwent isolated unilateral knee replacement surgery and no other joint replacement surgery. Patient-reported outcomes captured domains defined by the International Classification of Functioning, Disability, and Health. Domains of knee structure and function, activity limitation, and participation restriction were examined using growth-curve modeling over 5-year periods prior to and following surgery.
Preoperative worsening of the index knee was substantial in all domains of knee impairment, activity limitation, and societal participation. Pain intensity worsened only slightly from 5 years to 2.5 years prior to surgery, but worsened by ∼2 points (0–10-point scale) during the 2.5 years prior to surgery. Trajectories of improvement following surgery varied depending on the outcome measure. The contralateral knee also changed over time, such that by ∼2 years following surgery, pain was worse and by 3 years, activity limitation was worse in the contralateral knee as compared to the index knee.
Patients who elect to undergo knee replacement surgery demonstrate perioperative trajectories of change that influence most health domains for both the index knee and the contralateral knee. After a period of no change, escalation of pain and worsening functioning in the index knee begins ∼2.5 years prior to surgery, which may be a key trigger for surgery.
Moderate-to-severe symptomatic knee osteoarthritis (OA) affects multiple domains of a person's life and sometimes leads to knee replacement (KR) surgery ([1, 2]). Conceptualization of life domains has been most commonly described using the International Classification of Functioning, Disability, and Health (ICF) developed by the World Health Organization (). The ICF provides a hierarchical conceptual framework for health and a terminology for defining health dimensions. In the case of knee OA, the impaired body functions and structures domain includes knee pain, the most common reason for seeking KR surgery ([4-8]), and knee swelling and stiffness. The activity limitations domain describes person-level limitations in generic tasks, such as walking, stair climbing, and kneeling. The participation restrictions domain consists of adversely affected social life situations, such as recreational and occupational activities.
Davis and colleagues () among other investigators ([10-12]) have reported the trajectories of postoperative recovery of life domains following KR surgery. The literature generally suggests that after a rapid improvement in essentially all ICF domains during the first 6 months to 1 year following surgery ([9, 12]), the trajectory is essentially flat over subsequent years ([11, 12]). The contralateral limb gradually becomes more painful over the postoperative period, and by 3 years, it is more painful than the index (i.e., surgical) knee (). Evidence of the perioperative time course of change for both the surgical and nonsurgical limbs, however, is incomplete. The extent and rate of change in these same ICF-based domains in the years prior to surgery is unknown.
It is generally presumed that the index knees of potential candidates for KR surgery undergo gradual functional decline and worsening pain (). Given that OA is a chronic and gradually progressive disease and persons undergoing KR surgery generally report pain over a period of years (), this assumption of gradual worsening appears reasonable. However, we found no longitudinal studies that quantified the trajectories of preoperative worsening of various ICF domains combined with postoperative improvements in both the index and contralateral knees of patients undergoing KR surgery.
By knowing the “typical” trajectories of preoperative worsening and postoperative recovery in persons who eventually undergo KR surgery, both the patients and their physicians may make better-informed decisions regarding the timing and likely outcome of a potential KR surgery. Because preoperative scores are the strongest predictors of postoperative outcome and recovery ([11, 15]), knowing the type and rate of preoperative decline may assist in developing better interventions to reduce, stop, or reverse disease consequences earlier in the disease course. Knowing the type and extent of changes in the contralateral limb also could inform interventions designed to maintain or improve function, particularly given that the contralateral knee plays a major role in determining the extent of pain and function following KR surgery (). A comprehensive study of perioperative changes would inform patients and clinicians of the timing and consequences of disease progression associated with KR surgery for knee OA.
Our primary aim in the present study was to estimate the trajectories of change for a comprehensive set of ICF disablement model measures for both lower limbs up to 5 years prior to unilateral KR surgery and up to 5 years postoperatively. Specifically, we used data from the Osteoarthritis Initiative (OAI) () longitudinal cohort study to determine the trajectories of preoperative and postoperative pain, other knee-related symptoms, activity limitations, and participation restrictions. Because of limited longitudinally based evidence, particularly for the preoperative period, we used an exploratory rather than a confirmatory approach. For the 5-year preoperative period, evidence from nonarthroplasty patients suggested that we would find gradual worsening of the function and structure impairment and activity limitation domains for the index knees (), while for the contralateral knees, we expected flat trajectories. Postoperatively for these same ICF domains, evidence strongly suggested early substantial improvement during the first year and then a plateauing for the index knees, while for the contralateral knees, we anticipated gradual worsening (). For the participation restriction domain of the ICF, we suspected a steady worsening preoperatively and a more gradual worsening postoperatively, given that this domain is known to show gradual worsening in elderly persons ([18, 19]).
Data obtained from the OAI public-use data set (available online at http://oai.epi-ucsf.org) were used in the present study. The OAI is a prospective 5-year longitudinal cohort study of persons with, or at high risk of, knee OA. Subjects received no treatment as part of the study. The study was approved by the Institutional Review Board (IRB) of the OAI Coordinating Center at the University of California, San Francisco. The procedures followed were approved by the IRBs of each participating site. Subjects were recruited in similar proportions from 4 sites: the University of Maryland (Baltimore, MD), the Ohio State University (Columbus, OH), the University of Pittsburgh (Pittsburgh, PA), and Memorial Hospital of Rhode Island (Pawtucket, RI).
Exclusion criteria were the presence of rheumatoid arthritis, bilateral KR surgery or plans to undergo bilateral KR surgery in the next 3 years, bilateral end-stage radiographic knee OA, positive findings on a pregnancy test, use of ambulatory aids other than a single straight cane for >50% of the time, comorbid conditions that might interfere with participation, current participation in a randomized trial, and unwillingness to sign informed consent forms. In addition, because the study measured magnetic resonance imaging (MRI)–based changes, men weighing >130 kg and women weighing >114 kg were excluded because they were unable to undergo 3.0T MRI.
For the current study, data from persons undergoing unilateral KR surgery and no revision knee surgery or hip replacement surgery over the 5-year study period were included. There were 6 measurement sessions (baseline plus 1 per year) over the 5-year period, and each session was calculated as the number of days from surgery. Of a total of 216 persons who underwent KR surgery, 177 persons underwent isolated unilateral KR surgery. We excluded persons who underwent either bilateral KR or revision KR surgery (n = 35) because of the potential influence of repeated surgery on short-term and long-term recovery. We also excluded persons who underwent KR and hip replacement surgery (n = 4) because of the potential influence of hip surgery on the preoperative status and postoperative recovery.
We used the ICF to guide our choice of outcome variables. Knee-level measures are defined by the ICF as impairments in body system or organ-level structure or function, and we chose a knee pain rating in which knee pain at its worst over a 7-day period was measured on a 0–10 scale, with 0 indicating no pain and 10 indicating pain as bad as you can imagine. We also used the symptoms scale of the Knee Injury and Osteoarthritis Outcome Score (KOOS), a reliable and valid measure of 7-day knee-level symptoms that includes questions regarding the severity of knee swelling, grinding, catching, stiffness, and ability to straighten and bend the knee ([20, 21]). The KOOS symptoms scale is scored from 0 to 100, with 100 indicating no symptoms and 0 indicating severe symptoms. Both the pain measure and the KOOS symptoms measure for the index and contralateral knee were obtained for the OAI. The KOOS pain subscale ([20, 21]), a reliable and valid measure of function-related pain, was also used and is scored much like the KOOS symptoms scale. Participants completed all measures at home prior to visits to the OAI centers.
The activity limitations concept of the ICF was measured with the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) physical function scale, which asks the patient to rate the difficulty associated with 17 common activities, such as walking, sitting, and getting out of bed. The WOMAC physical function scale is a validated measure ([22-24]), with a range of scores from 0 (normal function) to 68 (severely affected function). We chose the WOMAC physical function scale as the most pure measure of activity limitation because the scale specifically does not inquire about symptoms associated with the activities, but only the extent of difficulty in doing each activity. Patients completed a WOMAC physical function scale questionnaire for each knee. To capture the participation restriction concept of the ICF, we used the Physical Activity Scale for the Elderly (PASE) (). The PASE is a validated measure of physical activity that measures the extent of purposeful activity, such as housework, home repair, yard work, recreation, and paid or volunteer work–related activities ([26, 27]). PASE scores increase as activity-related time increases. PASE scores were available for only the first 4 years of data collection, which allowed us to model up to 5 years of preoperative data, but only 4 years postoperatively. All ICF-based measures were obtained yearly over the study period.
A variety of covariate measures were used to assess for potential confounding. Comorbidity was measured using the validated Katz Comorbidity Scale (). Body mass index (BMI) was calculated using calibrated scales, and depressive symptoms were assessed with the validated Center for Center for Epidemiologic Studies Depression Scale (CES-D) (). All persons completed each measure on a yearly basis over the 5-year study period. Data were recorded in the OAI database to allow for determinations of the number of days from baseline to each followup visit and for the days from KR surgery to each visit.
Latent growth curve (LGC) modeling with individually varying times of observations and its 2-piece extension were used to estimate the trajectories of pain and functioning separately for the index and contralateral knees and for each ICF domain (). This modeling approach has several features: 1) models missing data through maximum likelihood estimator under missing at random and missing completely at random mechanisms, 2) estimates growth trajectories when the times of measurement vary between subjects, 3) estimates linear and nonlinear trends, and 4) specifies growth parameters (e.g., intercept, linear slope) as random coefficients, which allows for the incorporation of covariates in the estimation of trajectories (). The preoperative and postoperative periods were modeled separately.
The LGC with individually varying times of observations was used in an exploratory manner when the trend appeared smooth over time (e.g., postoperative PASE scores). Two-piece LGC modeling with individually varying times of observations was used when a particular trend appeared to have an elbow shape (e.g., preoperative WOMAC physical function scores for the contralateral knee).
The first step when exploring each trajectory was to start with the most complex model (e.g., intercept, linear slope, and quadratic slope factors). A model comparison approach was adopted to obtain the most parsimonious model for each ICF measure and for both the index knee and the contralateral knee. Consequently, all fixed and random growth parameters reported in this study were statistically significant (P < 0.05). The model-implied mean trajectories were then plotted against the observed data to assess the model fit. Mplus (version 6.11) software was used to fit the models.
Repeated measurements were regressed on sex, presurgical age, comorbidity, and BMI as time-invariant covariates, one at a time, within the LGC models to explore obvious omitted variables in estimating each trajectory. To eliminate all potential threats to inferential validity, all time-invariant and time-varying covariates should have been included in estimating the LGCs, but issues related to the sample size relative to the model complexity and collinearity issues with the time-varying covariates prevented us from using this approach.
During years 1, 2, 3, 4, and 5 of the study, a total of 19, 31, 38, 40, and 49 patients underwent KR surgery. The mean ± SD age of our study sample during the year of surgery was 67.1 ± 8.5 years; 58.2% of the patients were female, and 83.1% were Caucasian. The mean ± SD comorbidity score was 0.4 ± 0.8, and the mean ± SD BMI was 29.8 ± 4.7. The mean ± SD time from the baseline visit to KR surgery was 1,055 ± 502 days (range 16–1,819 days). A total of 94 persons had a left-sided KR and 83 had a right-sided KR, for a total of 177 KR surgeries in 177 patients. Additional study information is available online at http://www.sahp.vcu.edu/pt/about/riddle.html.
The most parsimonious models were used to estimate the trajectory of each ICF measure for the index knee and the contralateral knee, both preoperatively and postoperatively. For example, using 7-day levels of pain in the surgical knee, there was an increase in the mean pain scores during the years preceding surgery (Figure 1A). Modeling suggested that pain increased across the years (positive linear time estimate) and was accelerating prior to surgery (positive quadratic time estimate). The patients' mean level of pain increased by ∼2 points during the 2.5 years prior to surgery. The preoperative pain score for the contralateral knees had only 1 significant parameter, the intercept, while the WOMAC postoperative trajectory for the surgical knee had 3 significant growth parameters, the intercept, the linear slope, and the quadratic slope. Table 1 shows the growth parameter estimates for each model. Thus, the preoperative pain model for the contralateral knee is simply a flat line (Figure 1B), while the model for the WOMAC postoperative trajectory for the surgical knee demonstrates accelerated improvement until about 2 years after surgery (Figure 2A).
|Variable, knee, study period||Growth parameters|
|Intercept||Linear slope 1||Quadratic slope||Linear slope 2|
|WOMAC physical function|
From 5 years to ∼2.5 years prior to surgery, the index knee pain and KOOS symptoms and pain scale scores demonstrated minimal or no worsening (see Figures 1, 3, and 4). Accelerated worsening began ∼2 years prior to KR surgery. As seen in Figure 1, postoperative pain intensity changes in the index knee from a few months prior to surgery to a few months following surgery showed reductions on the order of 50%, on average. KOOS pain scale score changes were similar to knee pain scale scores (Figure 3). KOOS symptoms scale scores for the index knee improved by ∼25% over the same time period. Following surgery, the measures for the index knee demonstrated an initial improvement during the first 18 months postoperatively, followed by a plateauing of scores up to 5 years postoperatively. The contralateral knee demonstrated flat trajectories preoperatively, with pain and symptoms worsening in a linear manner during the 5 years following surgery. By ∼2 years following surgery, pain in the contralateral knee exceeded the pain in the index knee. Trajectories for both the index knee and the contralateral knee are illustrated in Figures 1, 3, and 4, and growth curve parameters are shown in Table 1.
For the index knee, the WOMAC physical function scale scores demonstrated essentially no change from preoperative years 5 through 2.5. Beginning ∼2.5 years prior to surgery, the WOMAC physical function scores worsened in a linear manner until surgery. Postoperative changes in these scores relative to the preoperative scores improved by ∼50% for the index knee (see Figure 2A and Table 1). Following surgery, the index knee improved for ∼3 years, after which the WOMAC score trajectories demonstrated mildly accelerated worsening. The contralateral knee demonstrated a plateauing of the WOMAC scores during preoperative years 5–2.5 and then a slightly linear worsening until prior to surgery. Early postoperative scores improved by ∼33% for the contralateral knee, and the WOMAC physical function scores worsened postoperatively in a linear manner to the extent that by ∼3 years following surgery, they exceeded the scores for the index knee.
The PASE scores worsened in a linear manner during the 5 years prior to surgery (see Figure 5 and Table 1). Early changes following surgery indicated worsening of the PASE scores on the order of 10% relative to the preoperative scores. During the 5 years following surgery, the PASE scores remained flat up to ∼4 years following surgery and stayed below the immediate preoperative levels.
Random growth parameters were regressed on age, sex, comorbidity, and BMI. From 14 LGC models involving 24 random growth parameters (14 for the intercept, 6 for the linear slope, and 4 for the quadratic slope) with 4 predictors each, we estimated 96 regression coefficients (i.e., 24 growth parameters × 4 covariates). Significant weights appeared to have small effect sizes. As no clear pattern of covariate effects emerged from the data, the results were reported without the covariates (see http://www.sahp.vcu.edu/pt/about/riddle.html for a table of regression coefficients for the covariates as well as additional study information).
Our study links multiyear preoperative changes with multiyear postoperative recovery for both the index knee and the contralateral knee in patients undergoing knee replacement surgery. Several novel findings emerged. Generally, we found that perioperative trajectories varied depending on knee surgery status and ICF domain. The most novel findings were related to changes during the preoperative period. Patients with knee OA typically demonstrate gradual worsening (); however, we found that clinically important worsening of the index knee was generally flat, with acceleration during the 2.5 years prior to surgery. As suspected, the contralateral knee showed either no change or only slight worsening, and the participation domain demonstrated steady worsening over the entire 5-year preoperative period. Despite substantial changes in the index knee, the preoperative period appears to be a stable period for the contralateral knee.
To our knowledge, our study is the first to describe preoperative trajectories of the full complement of ICF-based outcome measures up to 5 years prior to KR surgery for both index knees and contralateral knees. These data illustrate the disablement process in patients who ultimately elect to undergo major surgery for their arthritis. With regard to the index knees, our finding surprised us, in that patients appeared to experience progression with an essentially flat trajectory or with no change until ∼2.5 years prior to surgery. At this time, an onset of accelerated worsening of knee pain, knee symptoms, and function occurs. This 2.5-year period would appear to be extremely important, as it is likely the time during which patients make the decision to undergo KR surgery (). While our study was not designed to identify causes of worsening, evidence suggests the potential for a variety of different triggers. In some patients, these triggers may be related to physical factors, such as knee injury or weight gain ([33, 34]), or a variety of psychological and pain modulation factors related to worsening pain and function ([35-37]). Other patients undergoing KR may represent a phenotype of arthritis that is prone to worsening later in the disease process (), while for others, there may be reciprocal effects due to a multiplicity of factors.
The preoperative worsening we found in the index knees likely contributed to the patient's decision to undergo surgery, but it is also possible that once a patient decides to have KR surgery, a very emotionally difficult decision for some patients ([39, 40]), the decision itself may affect subsequent scores on self-report measures. For example, some patients may score higher (worse) on pain and function measures as a form of cognitive dissonance () to support the surgical decision. Artificially inflated scores may assist in supporting a patient's difficult decision to undergo surgery. While we found no evidence of inflated preoperative scores, potential triggers for preoperative worsening need to be explored. Studies that examine the causes of worsening could assist in identifying interventions designed to limit or prevent worsening at a time that may be critical for reducing the risk of future arthroplasty.
Improvements in index knee functions/structures and activity limitations ∼3 months after surgery were consistent with prior evidence ([11, 31]) and were on the order of 25–50%. Participation restriction showed mild worsening over the same time period, which conflicts with previous research (), but we suspect that at least some of this difference is attributable to the different measures used to quantify participation restriction. An ∼50% improvement in activity limitation scores for the index knee 1 year following surgery is highly consistent with results of previous studies ([11, 12, 31]). Functions/structures and activity limitations improved slightly in the contralateral knee during the immediate perioperative period, but as we suspected based on prior evidence of pain (), these domains showed steady worsening during the 5 years following surgery. This gradual worsening may be occurring in part because of both the high rate of bilateral knee OA in patients undergoing KR surgery and the patients' increasing age relative to the preoperative period. Up to 90% of patients undergoing KR surgery have contralateral knee OA ().
Postoperative recovery has been described frequently in the literature ([9-12, 43, 44]), but we found only 1 study () that examined a variety of ICF-based outcomes up to 5 years following surgery. Nilsdotter and colleagues () examined 6-month, 1-year, and 5-year outcomes in 102 patients following KR surgery. The investigators found that 5-year KOOS pain scores were not significantly different from the 1-year outcomes but that activity limitation outcomes at 5 years worsened slightly relative to the 1-year measures.
A major difference between our work and that of Nilsdotter and colleagues was that we reported findings for both the index knee and the contralateral knee. Nilsdotter et al reported person-level deficits in various ICF domains. The OAI investigators modified the KOOS by asking participants to provide responses for each knee. For example, in the KOOS symptoms scale, one question was, “Do you have swelling in your right knee?” By asking patients to focus on either the right or left knee, OAI data allowed for descriptions of the extent of deficits in body structure/function in each knee. In the case of knee symptoms, we believe this to be a reasonable approach because the questions ask about knee-specific problems.
The WOMAC physical function scale, however, may be problematic in this regard because the questions are about person-level problems, such as difficulty with walking, sitting, or standing. The strategy we and other investigators have used was to record the worse (more involved) WOMAC physical function score to describe person-level activity limitations ([45, 46]). In the current study, we chose to report both the index and contralateral knee scores to provide a comprehensive summary of activity limitations thought to be associated with each limb. However, this approach may have led to errors in the WOMAC estimates, particularly for the contralateral knee. Patients may not be able, for example, to determine the extent to which the right versus the left knee contributes to difficulty walking, particularly when both knees or hips are symptomatic. This is particularly relevant given that ∼30% of patients undergo hip or contralateral KR surgery within 5 years of surgery on the index knee ().
Given these limitations, our work confirms and extends the findings of Nilsdotter et al (). Five-year pain outcomes for the index knees approximate those obtained at 1 year, while activity limitation measures are worse at year 5. Our 5-year postoperative findings should be interpreted with caution because only 19 subjects had surgery during the first year of data collection and could therefore provide data for 5 years following surgery. Our data suggest that at least some of this worsening may be attributable to the contralateral knee.
Worsening of the contralateral knee following surgery has been described previously. Farquhar and Snyder-Mackler () studied 60 patients for 3 years following KR surgery and found that knee pain scores and quadriceps strength were worse in the contralateral knee as compared to the index knee. Our study extends these findings. Contralateral knee pain, other knee symptoms, and activity limitations all worsen steadily in the years following KR surgery, and this worsening begins a few months following surgery. By ∼3 years following surgery, the contralateral knee is, on average, both more painful and less functional than the index knee. Our study thus suggests that worsening of the contralateral knee is global, affecting function and other knee symptoms as well as pain. This comprehensive worsening of the contralateral knee persists for up to 5 years postoperatively.
Participation in socially expected activities demonstrated steady reductions during the 5-year preoperative period and then plateaued postoperatively, but at a level that was slightly lower than that during the immediate preoperative level. Reductions in PASE scores over time are typical in persons ages 60 years and older ([18, 25]). Average 5-year declines for persons in this age range were similar to those seen in our study. Following surgery, PASE scores remained essentially the same, which suggests that the KR surgery may have halted the decline in PASE scores typically seen for persons in this age range.
Patients and physicians may find this information useful when attempting to understand the path toward, and recovery following, KR surgery. Physicians treating patients who may be candidates for KR surgery may find the estimated trajectories helpful when assessing their patients, but it is important to note that ours is not a traditional prognostic study. We do not know, for example, whether persons who are treated medically rather than surgically but have similar levels of OA experience similar trajectories. The trajectories we report herein can provide patients with an indication of how their symptom and functional status compares to that of typical patients who eventually had KR surgery. Following KR surgery, the outcome trajectories can describe outcomes patients might expect for short-term and longer-term pain, knee function, and person-level and societal function.
The main limitation of our study is the relatively small sample size of persons at the tails of the study period, but particularly at 5 years postoperatively. The sample size also restricted our ability to fully account for covariates in the final models. Use of the WOMAC physical function scale to describe activity limitations in both the right and left limbs may be problematic because patients must determine the extent to which the right knee versus the left knee contributes to person-level difficulties in activities such as walking and stair climbing. In addition, because of the very limited evidence, particularly for the preoperative period, our analysis was exploratory in nature, and model-building was its emphasis. Future work using larger sample sizes should confirm or refute the trajectories we have reported herein.
In conclusion, we have described the perioperative trajectories of change for both the index knee and the contralateral knee in patients who eventually underwent KR surgery. Preoperative worsening, our novel and most important finding, was substantial for the index knee and occurred in an accelerated manner during the 2.5 years prior to surgery. The contralateral knee showed very little change leading up to surgery, but steadily worsened in most domains after surgery.
All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Riddle had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study conception and design. Riddle, Stratford, Jiranek, Dumenci.
Acquisition of data. Riddle.
Analysis and interpretation of data. Riddle, Perera, Stratford, Jiranek, Dumenci.