Underlying diagnosis predicts patient-reported outcomes after revision total knee arthroplasty

Abstract

Objective. To assess the association of underlying diagnosis with outcomes after revision total knee arthroplasty (TKA).

Methods. For this cohort study we used prospectively collected data from the Mayo Clinic Total Joint Registry on all revision TKA patients from 1993 to 2005 with 2- or 5-year response to a validated knee questionnaire that assesses pain and function. We used logistic regression to assess the odds of moderate–severe activities of daily living (ADL) limitations and moderate–severe index knee pain 2 and 5 years after revision TKA. Odds ratios (ORs) and 95% CIs are presented.

Results. The underlying diagnosis for the 2- and 5-year cohorts was loosening, wear or osteolysis in 73% and 75%; dislocation, bone or prosthesis fracture, instability or non-union in 17% and 15%; and failed prior arthroplasty with components removed or infection in 11% and 11%, respectively. In multivariable adjusted analyses that included preoperative status, compared with patients with loosening/wear/osteolysis, patients with dislocation/fracture/instability/non-union had an OR of 2.1 for moderate–severe ADL limitation (95% CI 1.3, 3.1, P < 0.001) and those with failed prior arthroplasty/infection had an OR of 1.1 (95% CI 0.6, 1.8, P = 0.4). At 5 years, differences were no longer significant. In multivariable adjusted analyses, compared with patients with loosening/wear/osteolysis, patients with dislocation/fracture/instability/non-union had an OR of 2.0 for moderate–severe pain (95% CI 1.3, 3.1, P < 0.01) at 2 years and an OR of 2.1 (95% CI 1.3, 3.8, P = 0.01) at 5 years. Failed prior arthroplasty/infection was not significantly different than the reference category.

Conclusion. Underlying diagnosis is independently associated with ADL limitations and pain after revision TKA. This information can help patients have realistic expectations of outcomes.

Introduction

Revision total knee arthroplasty (TKA), a common orthopaedic procedure, is performed for improvement of pain and function in patients who have previously had a primary TKA [1]. In the US Medicare population, per capita utilization of revision TKA increased 59.4% over 20 years (from 3.2 procedures per 10 000 Medicare enrolees in 1991 to 5.1 procedures per 10 000 in 2010) [2].

Most published literature has focused on predictors of outcomes in patients with primary TKA, with very few studies focusing on revision TKA. The rapidly increasing annual volume of revision TKA implies that we need to learn more about predictors of outcomes after revision TKA. Gender was not associated with function outcome in 67 revision TKAs at the 2-year follow-up [3]. Studies identified BMI ≥40, higher Deyo–Charlson score, female sex, and age >80 years as significant predictors of moderate–severe overall activity limitation after revision TKA [4] and that BMI was not associated with moderate–severe pain after revision TKA [5]. Other studies of revision TKA did not report the predictors of outcomes [6, 7]. One important potential predictor of outcomes is the underlying diagnosis for revision total hip arthroplasty (THA). To our knowledge, there are no studies to date have examined this association.

Our objective was to examine whether underlying diagnosis is associated with pain and function outcomes in revision THA patients. We hypothesized that the (i) underlying diagnosis of loosening, wear or osteolysis will be associated with better pain and function outcomes at both 2 and 5 years after revision THA compared with other diagnoses and (ii) the association of underlying diagnosis and post-TKA outcomes (pain and function) is mediated by preoperative status (preoperative pain and preoperative function, respectively).

Methods

We describe the methods and results as recommended in the Strengthening of Reporting in Observational Studies in Epidemiology (STROBE) statement [8]. This study was approved by the Mayo Clinic Institutional Review Board and all investigations were conducted in conformity with ethical principles of research.

Setting and participants

We used prospectively collected data from the Mayo Clinic Total Joint Registry, a large US institutional registry that collects data on every patient who undergoes arthroplasty at the Mayo Clinic, Rochester, MN, USA. At 2 and 5 years after arthroplasty, validated pain and function surveys are administered to patients at the clinic visit, by mail or on the telephone by trained registry staff. Several studies using data from these validated questionnaires have been published [4, 5, 9–11]. The Mayo knee questionnaire, a validated standardized questionnaire [11] that assesses knee pain and function, is administered to all patients who undergo TKA. A very high correlation (correlation coefficient 0.74) between mailed and in-person physician-administered Mayo knee questionnaires has been demonstrated [12]. This questionnaire is similar to the Knee Society scale [13]. Patients were included in this study if they underwent revision TKA between 1993 (start of electronic data capture) and 2005 and completed either a 2- or 5-year patient survey. The Mayo Clinic Institutional Review Board approved the study and waived the requirement for patient consent for this database study. All investigations were conducted in conformity with ethical principles of research.

Outcomes of interest

The Mayo knee questionnaire [12] was the source for the two patient-reported outcomes, overall moderate–severe activities of daily living (ADL) limitation and moderate–severe pain 2 or 5 years after revision TKA. Responses to limitations in three ADLs—walking, climbing stairs and rising from a chair—were categorized into no, mild, moderate or severe limitation for each activity, as previously [11]. The presence of two or more activities with moderate or severe limitation was classified as overall moderate–severe ADL limitation (reference, all other categories), as previously [4, 11]. The pain question is similar to the pain question in the Knee Society score, a validated TKA outcome instrument [14–16]: ‘Do you have pain in the knee in which the joint was replaced?’ The responses were no pain, mild (occasional), stairs only, walking and stairs (all combined into reference category); and moderate (occasional), moderate (continuous) and severe categories combined into the outcome variable moderate–severe pain. Most importantly, categorization of pain and ADL limitation outcomes was an a priori decision, based on recommendations from an experienced orthopaedic surgeon (D.G.L.). Moderate–severe ADL limitation or moderate–severe pain were designated as highly undesirable arthroplasty outcomes, similar to previous studies [17, 18].

Predictor of interest

Operative diagnosis was the main predictor of interest. As previously [4, 5, 18, 19], based on an a priori decision to avoid too many categories and as per an experienced orthopaedic surgeon (D.G.L.), diagnoses were categorized as (i) loosening, wear or osteolysis, (ii) dislocation, bone or prosthesis fracture, instability or non-union and (iii) failed prior arthroplasty with components removed or infection.

Covariates of interest

We included several covariates at the time of surgery in our analyses based on prior documented association with pain/function outcomes after TKA [4, 11] or important variables suspected to be potential confounders. These included (i) demographic variables, including age and gender; (ii) clinical variables, including BMI [20]; American Society of Anaesthesiologists (ASA) class [21, 22]; depression and anxiety based on the presence of International Classification of Diseases, 9th revision (ICD-9) codes; and medical co-morbidity assessed by the Deyo–Charlson index, a validated co-morbidity measure [23] consisting of 17 co-morbidities [24, 25] based on ICD-9 codes; (iii) distance from the medical centre and (iv) preoperative overall ADL limitation or preoperative pain. Age was categorized into ≤60, >60–70, >70–80 and >80 years; BMI into <25, 25–29.9, 30–34.9, 35–39.9 and ≥40; distance from the medical centre into 0–100 miles, >100–500 miles, >500 miles [9, 20, 26] and ASA class into I–II vs III–IV [17, 18], as previously.

Data sources

Anxiety, depression and Deyo–Charlson co-morbidities were obtained from linked Mayo Clinic electronic databases. Operative diagnosis, demographics, clinical variables, preoperative limitation in ADLs, preoperative pain and the distance from the medical centre were obtained from the Mayo Total Joint Registry.

Bias

We minimized confounding bias by including factors previously known or suspected to be associated with outcomes after TKA, but recognize that residual confounding is a limitation of cohort study design. We anticipated non-response at both 2 and 5 years, which is a study limitation.

Sample size

We did not perform any formal sample size calculations. We wanted a large enough sample to study without having too long a study period, therefore we chose all eligible patients from 1993 to 2005.

Statistical analyses

We used univariate and multivariable logistic regression models using a generalized estimating equations approach to adjust the standard errors for the correlation between observations on the same subject due to both knees having been replaced and/or multiple operations on the same knee. We assessed the association of the operative diagnosis with the odds of moderate–severe ADL limitation and moderate–severe pain at both 2 and 5 years after revision THA, analyses specified a priori. The multivariable models included age, gender, BMI, ASA class, anxiety, depression, Deyo–Charlson index and distance from the medical centre for all models and preoperative overall ADL limitation for the ADL limitation models or preoperative pain for the pain models. In exploratory analyses we examined each ADL limitation as an outcome. These are reported in supplementary tables available at Rheumatology Online. We reported odds ratios (ORs) with 95% CIs. A P-value <0.05 was considered statistically significant.

Results

The 2- and 5-year revision TKA cohorts had a mean age of 69 years each, 51% and 49% were female, and 20% each were <60 years of age. BMI was ≥40 kg/m² in 7% and 5% and ASA score was class III/IV in 50% and 47%, respectively. Anxiety was seen in 5% and 3% and depression in 8% and 6%, respectively (Table 1).

Table 1.

Clinical and demographic characteristics

	2-year (n = 1533)	5-year (n = 881)
Age, mean (s.d.), years	69 (10)	69 (10)
Men/women, %	49/51	51/49
Age groups, %
≤60 years	20	20
>60–70 years	29	31
>70–80 years	42	41
>80 years	9	8
BMI, %
<25 kg/m2	13	14
25–29.9 kg/m2	36	39
30–34.9 kg/m2	29	27
35–39.9 kg/m2	14	14
≥40 kg/m2	7	5
ASA class
Class I–II	50	53
Class III–IV	50	47
Deyo–Charlson index, mean (s.d.)	1.0 (1.7)	0.8 (1.4)
Psychological comorbidity, %
Anxiety	5	3
Depression	8	6
Operative diagnosis, %
Loosening/wear or osteolysis	57	61
Dislocation, bone or prosthesis fracture, instability, non-union	22	20
Failed prior arthroplasty with components removed or infection	21	19
Distance category, %
0–100 miles	32	34
>100–500 miles	57	54
>500 miles or non-US	11	12
Preoperative pain, %
None/mild	46	46
Moderate–severe	54	54
Preoperative overall ADL limitation, %
None/mild	12	10
Moderate–severe	88	90

Response rates for revision TKA cohorts were 57% (1533/2695) and 48% (881/1842) at 2 and 5 years, respectively (Supplementary Table S1, available at Rheumatology Online). Non-responders were more likely than responders to have underlying diagnoses of failed arthroplasty with components removed or infection, have slightly higher co-morbidity and were younger. Other characteristics, including sex, BMI, income category and distance from medical centre, were similar to the responders.

Unadjusted analyses: limitations in ADLs and pain

For the 2- and 5-year follow-ups, moderate–severe ADL limitations were reported by 42% and 51% with loosening/wear/osteolysis, 53% and 48% with dislocation/fracture/instability/non-union, and 53% and 52% with failed prior arthroplasty/infection, respectively (Supplementary Table S2, available at Rheumatology Online). Respective numbers for moderate–severe pain at 5 years were 19% and 22%, 30% and 32%, and 20% and 26% (Supplementary Table S2, available at Rheumatology Online). Limitations in walking, climbing stairs and getting up from a chair are shown in Supplementary Table S3, available at Rheumatology Online.

Compared with patients with loosening/wear/osteolysis, patients with dislocation/fracture/instability/non-union were 1.6 times as likely to report moderate–severe overall ADL limitation and 1.7 times more likely to report moderate–severe pain 2 years after revision TKA (Supplementary Table S2, available at Rheumatology Online); patients with failed prior arthroplasty with components removed or infection were 1.6 times more likely to report moderate–severe overall ADL limitation. At 5 years, differences were significant only for higher odds of moderate–severe pain in those with dislocation/fracture/instability/non-union (Supplementary Table S2, available at Rheumatology Online).

Multivariable adjusted analyses

In multivariable adjusted analyses that included nine-additional variables including preoperative ADL limitations, compared with patients with loosening/wear/osteolysis, patients with dislocation/fracture/instability/non-union had significantly higher odds (2.1 times) of moderate–severe ADL limitation at 2 years after revision TKA (Table 2). Other variables significantly associated with moderate–severe ADL limitations at 2 years were age 61–70 years (OR = 0.6; compared with age ≤ 60 years); BMI ≥ 40 (OR = 2.7); ASA class 3–4 (OR = 1.5); distance from the medical centre >500 miles/non-US residence (OR = 2.3); preoperative moderate–severe ADL limitation (OR = 3.3); and depression (OR = 2.6) (Supplementary Table S4, available at Rheumatology Online). At 5 years, differences by diagnosis in postoperative ADL limitations were no longer significant. Other variables significantly associated with moderate–severe ADL limitations at 2 years were female gender (OR = 2.1); higher Deyo–Charlson index (OR = 3.1 per 5-point increase); and preoperative moderate–severe ADL limitation (OR = 3.5) (Supplementary Table S4, available at Rheumatology Online).

Table 2.

Multivariable adjusted association of diagnosis with ADL limitation and pain after revision TKA

	2 years			5 years
	OR	95% CI	P-value	OR	95% CI	P-value
Moderate–severe ADL overall limitationa
Dislocation, bone or prosthesis fracture, instability, non-union	2.1*	(1.3, 3.1)	<0.001	0.8	(0.5, 1.4)	0.45
Failed prior arthroplasty with components removed or infection	1.1	(0.6, 1.8)	0.78	0.4	(0.2, 1.0)	0.05
Moderate–severe painb
Dislocation, bone or prosthesis fracture, instability, non-union	2.0*	(1.3, 3.1)	<0.01	2.1*	(1.3, 3.8)	0.01
Failed prior arthroplasty with components removed or infection	1.1	(0.7, 2.0)	0.64	0.6	(0.2, 1.9)	0.41

Reference category: loosening/wear or osteolysis. ^aAdjusted for nine covariates/confounders: age, gender, BMI, ASA class, anxiety, depression, Deyo–Charlson index, distance from the medical centre and overall preoperative limitation in three activities. ^bAdjusted for nine additional covariates/confounders: age, gender, BMI, ASA class, anxiety, depression, Deyo–Charlson co-morbidity score, distance from the medical centre and preoperative pain. *Significant ORs with P-values <0.05.

In multivariable adjusted analyses that included nine-additional variables including preoperative pain, compared with patients with loosening/wear/osteolysis, patients with dislocation/fracture/instability/non-union had significantly higher odds of moderate–severe ADL limitation at 2 and 5 years after revision TKA (OR = 2.1 for each). Failed prior arthroplasty with components removed/infection did not differ significantly from loosening/wear/osteolysis for ADL limitation or pain. Other variables significantly associated with moderate–severe pain at 2 years were age 61–70 years (OR = 0.5) and age 71–80 years (OR = 0.4); distance from the medical centre, 100–500 miles (OR = 2.8) and >500 miles/non-US residence (OR = 3.3); preoperative moderate–severe pain (OR = 1.8); and anxiety (OR = 3.7) (Supplementary Table S5, available at Rheumatology Online). Other variables significantly associated with moderate–severe pain at 5 years were distance from the medical centre, 100–500 miles (OR = 2.0) and >500 miles/non-US residence (OR = 2.7); and preoperative moderate–severe pain (OR = 2.4) (Supplementary Table S5, available at Rheumatology Online). Multivariable adjusted association of diagnosis with the limitation of each ADL is shown in Supplementary Table S6, available at Rheumatology Online.

Discussion

In this study we found that in univariate analyses the underlying diagnosis was associated with overall moderate–severe ADL limitation at 2 years and moderate–severe pain at both 2 and 5 years after revision TKA. After multivariable adjustment for variables including preoperative limitations and pain, the associations noted between diagnostic category and ADL limitations and pain remained significant at 2 and 5 years, with only one exception. Contrary to our hypothesis that associations will be attenuated with multivariable adjustment, adjustment led to even higher odds in multivariable adjusted compared with univariate analyses. The exception was that the association of a diagnosis of failed prior arthroplasty with components removed or infection with moderate–severe ADL limitation at 2 years was mediated by other covariates, including preoperative ADL limitation, since the OR noted in the univariate regression attenuated and was no longer significant in the multivariable models. Several findings in this study merit further discussion.

First, we found that an underlying diagnosis of dislocation/fracture/instability/non-union was associated with higher odds of moderate–severe ADL limitation 2 years after revision TKA compared with loosening/wear/osteolysis. This is a novel finding. Since these analyses were adjusted for preoperative ADL limitations (in addition to several other confounders), this difference is not attributable to differences in preoperative status (and other variables) between diagnostic categories. Patients with diagnoses of dislocation/fracture/instability/non-union may be more limited than those with loosening/wear/osteolysis in complying with post-TKA rehabilitation, which is associated with long-term functional outcome after arthroplasty [27].

We did not find any long-term studies of diagnoses and outcomes in revision TKA. A recent study found that functional improvements after inpatient rehabilitation (for an average of 10 days) were lower in patients with infection compared with mechanical or pain as underlying reasons for revision hip arthroplasty [28]. In contrast, we did not find any significant differences in ADL limitation or pain outcomes by this diagnosis in patients undergoing revision TKA. The differences may be due to differences in the joint (knee vs hip), time of assessment (2 and 5 years vs inpatient rehabilitation for 10 days) and outcome measures (specific ADLs vs Functional Independence Measure, a measure of physical and cognitive disability). Nevertheless, more studies addressing these aspects of revision TKA are needed.

Second, we found that compared with patients with loosening/wear/osteolysis, diagnosis of dislocation/fracture/instability/non-union was associated with significantly higher odds of 1.7 times each of moderate–severe pain 2 and 5 years after revision TKA in unadjusted analyses and 2 and 2.1 times odds, respectively, in multivariable adjusted analyses. In absolute terms, 30% and 32% of patients with dislocation/fracture/instability/non-union had moderate–severe pain at 2 and 5 years after revision TKA. A diagnosis of dislocation/fracture/instability/non-union may be associated with lower compliance with post-THA rehabilitation, given a more challenging early postoperative course for these patients and possibly higher rates of muscle atrophy prior to surgery. This may lead to suboptimal pain relief. Preoperative pain is unlikely to explain these findings, since these associations did not attenuate, but became stronger when analyses were adjusted for preoperative pain and other demographic and clinical variables. This is a novel finding that has significant implications. This information should be incorporated into informed consent discussions as well as help the patient have more realistic expectations for revision TKA. Thus our results add to the current literature.

Our study has several strengths and limitations. Study strengths include a large sample size that avoids the problem of type II error (missing a significant finding when one exits due to underpowered small sample sizes), analyses of prospectively collected data, examination of both pain and ADL limitations at 2 and 5 years, performance of multivariable adjusted analyses accounting for multiple covariates and confounders, and robustness of our findings. Our study has several limitations. Despite our attempt to control for multiple factors, residual confounding is possible due to the non-randomized cohort study design. Our study was performed at a single centre, therefore findings may not be generalizable to all settings. However, the clinical and demographic characteristics of our cohort are similar to other studies of revision THA. Non-response may have biased our results towards or away from the null, since we do not know whether subjects with post-TKA outcomes had similar chances of non-response for each category of diagnosis. The response rates are similar to those for large surveys of this size, reported at a mean of 60% [29]. The non-response rate was higher at 5 years, making them potentially more biased.

In summary, in this study we found that an underlying diagnosis of dislocation/fracture/instability/non-union was associated with a higher risk of moderate–severe ADL limitation at 2 years and a higher risk of moderate–severe pain at 2 and 5 years. This information can be used to better inform patients prior to revision THA, regarding their risk of ADL limitation and pain. Future studies need to examine the underlying mechanism of the higher risk of suboptimal outcomes in patients with dislocation/fracture/instability/non-union undergoing revision THA.

Rheumatology key messages.

• Underlying diagnosis of dislocation/fracture/instability/non-union was associated with more activity limitations after revision total knee arthroplasty.

• Patients with dislocation/fracture/instability/non-union had higher risk of moderate–severe index arthroplasty pain after revision total knee arthroplasty.

Supplementary data

Supplementary data are available at Rheumatology Online.