Background
Total knee arthroplasty (TKA) is an extremely effective surgical treatment option for patients with end-stage refractory knee pain and associated functional disability. National Inpatient Sample (NIS) reported that 719,000 primary TKAs were performed in the U.S. in 2010 [
1], consistent with the projected estimates from a population-based study from the Olmsted County, Minnesota [
2]. The estimated cost per TKA is approximately US $24,200 [
3], leading to an annual cost burden of 17.6 billion for TKA in the U.S in 2010. TKA volume has increased dramatically in the last few decades [
2,
4] and is projected to reach 3.5 million TKAs annually by 2030 [
5].
Many studies have focused on time-trends in utilization and specific in-hospital or immediate post-arthroplasty outcomes [
6‐
9]. To our knowledge, there are no studies assessing whether there are any appreciable time-trends in post-operative PROs, i.e., if the PROs after TKA have improved, worsened or remained stable over time. This is a critical knowledge gap since TKA is an elective surgery that results in an impressive improvement in patient-reported outcomes (PROs), including pain, function and quality of life in most patients [
10]. A small but sizable proportion of patients have suboptimal pain (7-13%) and function (21-27%) outcomes post-TKA [
11,
12]. Even a 1% absolute increase means an additional 7,190 patients with suboptimal outcomes annually in the U.S. Socio-demographic and clinical characteristics of patients undergoing primary TKA cohort are changing rapidly with an increasing patient complexity over time in the U.S. [
13]. If there is an evolution in PRO outcomes post-arthroplasty over time, we need to know and understand that from a policy maker, clinician and patient perspective. Therefore, the objectives of this study were to: (1) examine the time trends in preoperative PROs of functional outcome and pain in patients undergoing primary TKA; (2) assess whether the risk of suboptimal PROs post-TKA has changed over time; (3) assess whether the risk of improving or worsening PROs post-TKA over time is independent of the preoperative PROs; and (4) study how these time-trends in post-TKA outcomes impact patients in different categories of age, medical and psychological comorbidity and body mass index (BMI). We hypothesized that post-TKA PROs would improve over time (hypothesis 1) and that most time-trends in improving PROs will be seen in the oldest, and in those with lower comorbidity and lower BMI (hypotheses 2-4).
Methods
We followed the recommendations from the Strengthening of Reporting in Observational studies in Epidemiology (STROBE) statement [
14] to report study methods and results.
Study setting, data source and study eligibility
We included all patients from the Mayo Clinic Total Joint Registry who had undergone primary TKA between 1993-2005 and had completed a preoperative and/or 2-year postoperative pain and function questionnaire. The Mayo Clinic Total Joint Registry captures data on all patients who undergo knee, hip or other joint arthroplasty. For TKA, data are captured using the Mayo Knee Survey that has construct validity and reproducibility [
15]. Mayo Knee Survey has questions related to pain and function, similar to the validated American Knee Society Scale [
16], the most commonly used instrument in knee arthroplasty studies [
17]. Several previous studies using the Mayo Knee Survey have been reported [
12,
18‐
20]. Mayo Clinic’s Institutional Review Board approved this study.
Predictor of interest
The main predictor of interest was time-period, divided into 1993-1995, 1996-1998, 1999-2001 and 2002-2005. Given that our objective was to assess whether the pain and activity limitation had changed over time, we compared the prevalence of moderate-severe activity limitation and moderate-severe pain in the first period, i.e., 1993-95 to the last period, i.e., 2002-05. Since the entire study period (13 years) was not an exact multiple of 3, the last period consisted of 4-years.
Outcomes of interest
The main outcomes of interest were PROs, moderate-severe limitation of activities of daily living (ADL) and moderate-severe pain. For each of the 3 ADLs, walking, climbing stairs and getting out of a chair, patient responses were categorized into mild, moderate or severe limitation category, as previously [
12,
19,
20]. Presence of ≥2 ADLs with moderate or severe limitation was classified as overall moderate-severe activity limitation
a priori (ref, all other categories), as previously [
12,
18‐
21]. We categorized responses to the pain question “Do you have pain in the knee in which the joint was replaced? (Please mark only one answer.)” into reference category no/mild pain (no pain, mild pain, pain with stairs only, pain with walking and stairs) or moderate-severe pain (moderate occasional, moderate continuous or severe pain),
a priori, as previously [
18,
20,
22].
Covariates
We included several covariates (some potential confounders) that are associated with pain/function after TKA [
23‐
26], namely: demographics (age, sex); BMI; preoperative ADL limitation; preoperative pain; and medical and psychological comorbidity. Preoperative ADL limitation and pain were assessed by questions similar to the 2-year evaluation. Medical comorbidity was assessed using validated Deyo-Charlson index [
27], a weighted scale of 17 comorbidities (including cardiac, pulmonary, renal, hepatic disease, diabetes, cancer, HIV and so on), expressed as a summative score where a higher score indicates more comorbidity. We examined psychological comorbidity by the presence of anxiety and depression at the time of surgery, based on the presence of International Classification of Diseases, ninth revision (ICD-9) codes in the Mayo Clinic electronic databases.
Bias and sample size
We included several known confounders of function and pain outcomes to reduce confounding bias, but the cohort study design raises the possibility of residual confounding bias. We accounted for non-response bias, an expected limitation of this study, by presenting the non-responder characteristics and considering them in study interpretation. We did not perform any formal sample size calculations, since we anticipated >6,000 patients at both baseline and 2-year follow-up for our study, to provide adequate number of outcomes. Correlation of observations (bilateral TKA in a patient, simultaneous or sequential) was accounted for by using appropriate statistical methods.
Statistical analyses
We compared crude rates of moderate-severe ADL limitation and pain across the four pre-specified time periods using chi-square tests. We used univariate and multivariable-adjusted logistic regression models, using a generalized estimating equations (GEE) approach, to assess the association between the most recent time period (i.e., 2002-05 compared to 1992-95) and moderate-severe ADL limitation/pain 2-years after primary TKA. This method adjusted for the correlation between observations on the same subject due to both knees having been replaced. As specified a priori, we compared the first and the last time-periods. The multivariable models were hierarchical and included the variables: (1) model 1: age and sex; (2) model 2: model 1+ medical comorbidity; (3) model 3: model 2 + anxiety + depression; (4) model 4: model 3 + medical comorbidity; (5) model 5: model 4 + preoperative pain/ADL limitation (in the respective model); (5) model 6: model 5 + BMI.
Subgroup analyses were done by examining multivariable-adjusted model 5 in patient subpopulations of: (1) age, <65 vs. ≥65; (2) psychological morbidity, defined as presence of anxiety and/or depression (none vs. any); (3) BMI, <30 vs. ≥30; and (4) Deyo-Charlson index (zero vs. one or higher). Odds ratios (OR), 95% confidence intervals (CI) and p-values were calculated. A p-value <0.05 was considered statistically significant.
Discussion
We found a significant increase in the odds of suboptimal post-primary TKA function and pain outcomes from 1992-95 to 2002-05. In contrast, preoperative moderate-severe ADL limitation decreased and preoperative moderate-severe pain minimally increased from 1993-95 to 2002-05. The adjusted odds of 2-year post-TKA moderate-severe ADL limitation increased by 34% over the study period, while the odds of 2-year post-TKA moderate-severe pain increased by 79%. The absolute increases in post-TKA overall moderate-severe ADL and pain limitations between the first and last time-periods were 4.7% (19% vs. 23.7%) and 3.6% (5% vs. 8.6%), statistically significant and clinically meaningful. These findings add to the current knowledge in post-arthroplasty PROs. We are not aware of any other study of time-trends in PROs after primary TKA.
Our study addressed our first hypothesis, i.e., time-trends in post-TKA function and pain outcomes, which contrary to our expectation, actually worsened over time. Medical [
25,
28,
29] and psychological comorbidity [
30‐
32] and obesity [
33], factors associated with more complications and/or worse PROs after primary TKA, have worsened over time [
6,
13] in patients undergoing primary TKA. Interestingly, adjusting for preoperative medical and psychological comorbidity, age, BMI and other confounders (sex and preoperative pain and ADL limitation) had no/minimal effect on time-trends in post-TKA moderate-severe ADL limitation or pain. This implies that increasing comorbidity, BMI, preoperative pain and ADL limitation over time do not explain the time-related worsening in post-TKA ADL limitations and pain outcomes. An impressive finding was an almost doubling of moderate-severe pain from 5% to 8.5% over the 13-year study period. This increase in the proportion of patients with suboptimal pain outcomes 2-year post-primary TKA over 13-years was an unexpected finding that needs further study. This 8.5% annual rate translates into 61,115 moderate-severely painful TKAs in 2010 alone, which represents a significant public health and patient burden.
So, why in the face of lower proportion of patients with moderate-severe ADL limitation (albeit higher proportion with moderate-severe pain) preoperatively, are postoperative pain and function outcomes worsening over time? Since our final models were adjusted for preoperative status and various other important factors, the time-trends in post-operative pain and function are in fact independent of changes in preoperative pain and function, respectively. We think that the increase in odds of moderate-severe ADL limitation over time may be related to shorter inpatient stay and a lower proportion discharged to home over time [
6], both of which may lead to more suboptimal recovery; on the other hand, recent improvements in pre- and post-operative rehabilitation programs might be expected to improve post-TKA functional outcomes [
34]. Potential causes for worsening pain outcomes over time may be the increasing prevalence of osteoarthritis over time [
35,
36], which may lead to higher likelihood of other lower extremity joint involvement or the spine, which may contribute to persistent index TKA pain due to referred pain from another osteoarthritic joint (i.e. hip or back) or due to interference with optimal physical therapy and rehabilitation post-TKA [
37,
38]. The increase in odds of 1.34 in moderate-severe ADL limitation was more modest as compared to the increase in odds of moderate-severe pain (OR, 1.77).
Future studies need to examine whether other factors such as psychological distress (e.g. post-traumatic stress disorder), coping strategies, family support, pain medication use in the postoperative period and/or type of rehabilitation, can explain the worsening pain and function outcomes after primary TKA over time. We used a diagnostic code, not a questionnaire, for depression/anxiety and therefore we may have missed subclinical or undiagnosed disease. Severity of depression or anxiety was not measured in our study, which may also contribute to worsening of persistent pain/functional limitation post-TKA. Over time, TKA has been made available to a wider patient population, some of whom may have been excluded previously, due to an unfavorable patient profile, not likely to benefit optimally from TKA. Worsening in other patient characteristics (proportion with fibromyalgia, catastrophizing etc.) over time not measured in our study may also be contributing to worsening outcomes over time.
Our subgroup analyses provided important insights. We hypothesized that post-TKA pain and function outcomes would improve over time in the oldest, and in those with lower comorbidity and lower BMI. We noted worse pain and function outcomes in 2002-05 compared to 1993-95 in the patients 65 years and older and those without psychological comorbidity. This was contrary to our expectations, since we hypothesized that patients with psychological comorbidity and those in the younger age group (who also have worse psychological comorbidity) might be the ones at highest risk of worsening PRO outcomes over time. These associations were also independent of sex, preoperative pain/function, BMI and preoperative comorbidity. To our knowledge, this is the first study to report these findings. Another novel finding was that pain outcomes in the 2002-05 were worse compared to 1993-95 in patients with Deyo-Charlson index of one or more or BMI ≥30. Thus, our study identified specific patient populations who demonstrate worsening PRO outcomes post-TKA. Patients with these characteristics should be the focus of the future studies to investigate the reasons for worsening PROs over time and design strategies to improve PROs.
Study findings must be interpreted considering the study limitations. Non-response may have biased our results. Non-responders to 2-year survey were more likely than responders to have slightly higher co-morbidity, higher ASA class, greater distance to the medical center and were younger. However, the direction of bias is unclear since it’s not known how the patient characteristics associated with non-response might have impacted these time-trends. This is single center study and generalizability to other settings may be challenging. The similarity of our cohort to other published cohorts [
39‐
41] as well as a national U.S. sample [
7], supports the representativeness of our sample. Use of diagnostic codes for medical and surgical comorbidity may have led to under-recognition of these conditions, which would bias study results towards null, making our estimates conservative. Deyo-Charlson index that uses diagnostic codes is a validated comorbidity measure [
27] and the prevalence of depression using codes is similar to the 9-15% reported in studies using validated instruments for depression [
42‐
44]. The use of multi-modal pain protocols, in- and outpatient rehabilitation programs, practice of earlier joint mobilization as well as prosthesis and surgery techniques may have evolved over time and could have positively or negatively impacted pain outcomes, especially in the last few years. We could not explore these trends in the more recent as well as assess the contribution of these potential secular trends due to limited resources and lack of availability of these data in our databases. This should be explored in future studies. Our study had several strengths. We used a large sample from a Total Joint Registry, and adjusted analyses for potential confounders and covariates of pain and functional outcomes.
Grant support
JAS is supported by grants from the Agency for Health Quality and Research Center for Education and Research on Therapeutics (AHRQ CERTs) U19 HS021110, National Institute of Arthritis, Musculoskeletal and Skin Diseases (NIAMS) P50 AR060772 and U34 AR062891, National Institute of Aging (NIA) U01 AG018947, and National Cancer Institute (NCI) U10 CA149950, and research contract CE-1304-6631 from Patient Centered Outcomes Research Institute (PCORI). JAS is also supported by the resources and the use of facilities at the VA Medical Center at Birmingham, Alabama, USA.
Competing interests
There are no financial conflicts related directly to this study. J.A.S. has received research and travel grants from Takeda and Savient; and consultant fees from Savient, Takeda, Regeneron AND Allergan, and Novartis. D.G.L. has received royalties/speaker fees from Zimmer, Orthosonic and Osteotech, has been a paid consultant and owns stock in Pipeline Biomedical; his institution has received research funds from DePuy, Stryker, Biomet and Zimmer.
Authors’ contributions
JAS designed the study, developed the protocol and obtained IRB approval, conducted and interpreted the analyses, wrote the first draft of the manuscript and revised it and made the decision to submit it. DGL reviewed the protocol and suggested modifications, interpreted the analyses, critically revised the manuscript and approved the final version of the manuscript for submission.