Clinicians have been reporting about 20% poor outcomes in knee replacements for decades. This report demonstrated for the first time with a scientifically defensible method that they were right all along. This article also discuss why some people have good results and why some don’t.
Janis C. Kelly
January 31, 2019
Two characteristics that can be assessed before knee arthroplasty (KA) — pain catastrophizing and a larger number of painful body regions — are significant predictors of poor pain and function outcomes and might be useful for identifying patients for intervention prior to surgery, a study has found.
Low income was associated with poor pain outcomes after KA, whereas baseline pain and younger age predicted poor function outcomes.
The study by Levent Dumenci, PhD, from the Department of Epidemiology and Biostatistics, Temple University, Philadelphia, Pennsylvania, and colleagues was published online January 17 in Osteoarthritis and Cartilage.
Poor outcomes after total KA for osteoarthritis affect 200,000 patients per year in the United States, which represents about 20% of those who undergo this procedure.
Dumenci told Medscape Medical News, “Two key findings were surprising. First, clinicians have been reporting about 20% poor outcomes for decades. We demonstrated for the first time with a scientifically defensible method that they were right all along. Second, I have been developing and using mixture models for a few decades. Except for simulated data, I have never seen such a huge difference. The two groups that we are reporting are really different in kind.”
This study is a secondary analysis of data from the three-arm Knee Arthroplasty Skills Training (KASTPain) randomized clinical trial that compared outcomes of physical therapist–delivered pain coping skills training, arthritis education, and usual care for patients who had undergone KA. Outcome measures were the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) Pain and Physical Function Scales. According to the authors, none of the three interventions produced significant improvements in the WOMAC Pain or Function scores or in other secondary outcomes.
For the secondary analysis, the researchers combined the three treatment groups. Baseline variables that were assessed for ability to predict outcome trajectory type (good or poor) included WOMAC Pain or Function scores, arthritis self-efficacy, pain catastrophizing, painful body regions, depressive symptoms, age, sex, income, body mass index, comorbidity, and opioid use. Race/ethnicity was dichotomized to black vs all other.
The main objective of the secondary analysis was to identify clinically useful trajectories of pain and function recovery in persons following KA. An additional goal was to identify sets of presurgical variables that could predict subgroups likely to have poor pain and function trajectories.
Baseline characteristics included median knee pain duration of 6 years, mean WOMAC Pain score of 11.4, and mean WOMAC Function score of 37.2.
The researchers assessed pain and function 2 weeks before surgery and 2, 6, and 12 months after surgery in 384 participants. The researchers used piecewise latent class growth models to estimate pain and function trajectories and presurgery variables to predict trajectory types. They used a statistical modeling–based approach to determine outcome categories.
The researchers hypothesized that there would be about 20% poor outcomes predictable by pain and function trajectory–based classification and that the chance-corrected agreement between pain and function outcome types would predict which persons would have the same outcome for both pain and function. The authors also hypothesized that the chance-corrected agreement would be relatively high between pain- and function-based trajectory groupings.
As expected, WOMAC Pain and Function trajectories showed good outcomes in 82% of patients. These patients generally had high pain and poor function 2 weeks prior to surgery, a sharp drop in pain and improvement in function after surgery, and further gradual improvements in pain and function over the following year.
WOMAC Pain and Function trajectories showed poor pain and function outcomes in 18% of the sample. Patients with poor pain outcomes had slightly higher pain scores prior to surgery, a modest improvement in the 2 months after surgery, but no further improvement after that. Those with poor function outcomes had worse functional deficits before surgery, modest improvement in the 2 months after surgery, and no further improvement thereafter.
Most patients had similar trajectory types for pain and function, but about 9% had good/poor trajectory combinations.
Accounting for all other predictors, the researchers found that lower income, higher pain catastrophizing, and larger number of painful body regions were associated with increased risk for poor pain outcomes. Higher baseline pain, younger age, higher pain catastrophizing, and larger number of painful body regions were associated with increased risk for poor function outcomes.
The differences between the good and poor outcome trajectories were stark. For patients in the poor WOMAC Function and poor WOMAC Pain outcome subgroups, average 12-month outcome scores were similar to preoperative scores for those in the good pain and function outcomes subgroups, indicating substantial pain and poor function. The authors note that the mean WOMAC Pain score of approximately 10 at 1 year in the poor WOMAC Pain subgroup reflects substantial pain during many routine activities. It is equivalent to scores of moderate pain for all five WOMAC Pain items (walking, stair climbing, at night, sitting, and standing) or of severe pain on at least three of five items and mild pain on one.
“Because the poor outcome subgroups had persistent and substantial pain and poor function, methods for identifying these subgroups should be a high priority for clinicians treating these patients. It is possible that interventions targeting the factors associated with the poor outcome subgroups could make a substantial impact on the outcomes following surgery,” the authors write.
Unfortunately, the randomized controlled trial’s attempt to reduce pain catastrophizing failed. “While our KASTPain trial indicated no effect of pain coping skills training on outcome following KA for persons with pain catastrophizing scores of ≥ 16, the current analyses suggest that persons with extremely high levels of pain catastrophizing are at risk for poor outcome. This relatively small subgroup of patients may benefit from pain coping treatment though this requires further study,” the researchers explain. The primary KASTPain study results have not yet been published.
From the clinical perspective, Dumenci said, “Variables outside the outcomes of pain and function are useful in describing the poor and good outcome subpopulations, but it is my personal view that they are of little or no use when it comes to intervention. As the risk prediction literature testify, prediction models using such variables (age, race, catastrophizing, etc) do only a slightly better job than random selection (chance). That’s why clinicians do not trust them and do not use them in their practices. I do not blame them for that.
“If were a clinician, I would not change anything right now. But if I could determine, for example, that a particular patient has a 92% change of ending up with a poor outcome 1 week prior to a scheduled surgery or 3 weeks after the surgery with 94% accuracy using a scientifically defensible methodology, I would seriously consider my options for intervention,” Dumenci added.
The study was supported by grants from the National Institute of Arthritis and Musculoskeletal and Skin Diseases of the National Institutes of Health and from the National Center for Advancing Translational Sciences to Virginia Commonwealth University. The authors have disclosed no relevant financial relationships.
Osteoarthritis Cartilage. Published online January 17, 2019. Abstract