Rheumatoid Arthritis Flares after Total Hip and Total Knee Arthroplasty: Outcomes at 1 year

Abstract

Objective:

Most RA patients undergoing total hip(THA) and total knee arthroplasty(TKA) have active RA and report post-operative flares; whether RA disease activity or flares increase risk of worse pain and function scores 1 year later is unknown.

Methods:

RA patients were enrolled before THA/TKA. Patient Reported Outcomes (PROS) including Hip and Knee Osteoarthritis/disability and injury Outcome Scores (HOOS/KOOS) and MD assessments of disease characteristics and activity (DAS 28, CDAI) were collected before surgery. PROS were repeated at 1 year. Post-operative flares were identified using RA Flare Questionnaire weekly for 6 weeks, and defined by concordance between patient report plus MD assessment. We compared baseline characteristics and HOOS/KOOS scores using two-sample t-test/Wilcoxon rank-sum and Chi-squared/Fisher’s exact tests. We used multivariate linear and logistic regression to determine association of baseline characteristics, disease activity, and flares, with 1-year outcomes.

Results:

One-year HOOS/KOOS scores were available for 122 patients (56THA/66TKA). Although HOOS/KOOS pain was worse for patients who flared within six weeks of surgery; absolute improvement was not different. In multivariable models, baseline DAS28 predicted 1-year HOOS/KOOS pain and function; each 1 unit increase in DAS28 worsened 1-year pain by 2.41 (SE=1.05, p=.02) and 1-year function by 4.96 (SE=1.17, p=.0001). Post-operativeerative flares were not independent risk factors for pain or function scores.

Conclusion:

Higher disease activity increased risk of worse pain and function one year after arthroplasty, post-operative flares did not.

Utilization rates of total hip (THA) and total knee arthroplasty (TKA) remain high for patients with rheumatoid arthritis (RA), and the data guiding optimal management of RA medications during surgery is limited. The American College of Rheumatology (ACR) and the American Association of Hip and Knee Surgeons (AAHKS) guideline recommends withholding biologics and limiting glucocorticoids perioperatively (1). This recommendation weighed the risk of flares with concern for infection at the time of surgery, as DMARDs, biologics, and glucocorticoids are known to increase infection risk and withdrawal precipitates flares in other settings (1–4). While a causal relationship between biologics and infection after THA/TKA has not been conclusively established (5, 6), a patient panel supported this and concurred that risk of infection was much more important to them than post-operative flares, which they felt was an expected part of arthroplasty recuperation (7).

Optimizing the perioperative management of RA may improve long term arthroplasty outcomes, which are worse in patients with RA compared to those with OA. Despite similar absolute improvements, HOOS function is worse (82.9 vs, 91.8, p<0.001) and the risk of a poor function (HOOS function <60) is four times higher compared to OA. In addition, four times more patients with RA have worse pain scores two years after surgery (defined as HOOS pain<60) (RA 12% vs. OA 3%, p<0.001) (8–10). The reasons for this are not known, but disease activity or flares may play a role.

The majority of patients undergoing THA and TKA are on biologics, methotrexate, or other DMARDs, yet disease activity at the time of surgery remains high (11). We have previously reported that the mean DAS-28 was 3.7 +/- 1.5 at the time of surgery, and only 30% of the cohort was in low disease activity or remission by DAS-28 criteria (12). Moreover, 65% of patients report a flare, with severe worsening of symptoms, within six weeks of THA and TKA.(11, 12) It is not known whether post-operative flares, or high disease activity affect 1 year surgical outcomes. The purpose of this study is to describe the relationship of peri-operative disease activity and flares of RA with pain and function scores one year after THA/TKA.

Materials and Methods:

The study is reported per the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) checklist for cohort studies (13). This study was approved by the hospital ethical review board (#2014–233) and all included patients signed informed consent

The RA Perioperative Flare Study is a prospective observational cohort study of patients with RA undergoing THA and TKA at a single, high-volume tertiary care center for musculoskeletal diseases November 2014 through April 2018. The records of all patients undertaking THA and TKA were screened to identify RA patients, and all patients with confirmed RA over the age of 18 were eligible. The first 120 patients in the cohort were recruited consecutively prior to surgery, either on the day of surgery or during the pre-operative visit, and the subsequent cases were preferentially selected from patients with positive anti-citrullinated peptide antibodies (ACPA). Participants were undergoing elective THA or TKA and met classification criteria for RA using ACR/EULAR) 2010 or 1987 criteria, or were confirmed as RA by the principal investigator (PI)(SG) so all consecutive patients including those with low disease activity could be included in the initial cohort.

Data collection:

Enrolled patients underwent a comprehensive evaluation including patient and physician reported measures and laboratory evaluations at baseline and at 6 weeks after surgery (Appendix Table 1). Baseline data were collected during the pre-op medical screening performed within 2 weeks of surgery, or on the day of surgery prior to the start of the operation.

Participants answered a questionnaire each week for 6 consecutive weeks post-operative addressing RA status and disease flare. Patient reported outcome measures (PROMs) were repeated at 1 year.

Baseline data collected included age, sex, comorbidities, duration of disease, and medications. Serologic status for rheumatoid factor (RF) and ACPA, and inflammatory markers were obtained at baseline prior to surgery. Surveys included the Routine Assessment of Patient Index Data (RAPID3), RA Disease Activity Index (RADAI), Clinical Disease Activity Index (CDAI), MD Health Assessment Questionnaire (MDHAQ) and Disease Activity Scale (DAS28-ESR) (14, 15). Joint counts (excluding the operated joint), inflammatory markers, DAS28-ESR, MDHAQ, CDAI, RADAI, and RAPID3 were completed at baseline and 6 weeks. Participants also completed the tender/painful joint region score modified from the RADAI questionnaire including a self-report of specific painful and swollen joints on a homunculus depicting 40 joints. The CDAI is a composite disease activity measure of TJC, SCJ, patient global and physician global assessments of disease activity scored 0–76, higher worse, with categories of high disease activity> 22, moderate 10.1–22.0, low 2.9–10.0, and remission ≤ 2.8)(15). The RAPID3 includes the 3 areas of the RA core set, physical function, pain, and a patient global estimate of disease activity, and is scored 0–10 with categories of high (> 4), moderate (2.01–4), low (1.01–2), and near-remission (≤1)(16, 17). The DAS28-ESR is a composite disease activity measure, scored 0–10, with activity categories of high (> 5.1), moderate (3.21–5.1), low (2.61–3.2), and remission (≤2.6); a change in score of 1.2 is considered significant for those with moderate disease activity (15, 17–21). The MDHAQ adds 2 function questions to the Health Assessment Questionnaire (HAQ), and is scored 0–3(16, 17, 22). Participants completed the Hip disability and Osteoarthritis Outcome Score (HOOS) or Knee injury and Osteoarthritis Outcome Scores (KOOS) at baseline and at 1 year. The HOOS and KOOS are validated lower extremity instruments that are sensitive to change after THA and TKA; in this study, HOOS/KOOS function refers to activities of daily living (ADL) function, which is relevant after THA/TKA(23). The mean clinically important difference is 10 points on a 1–100 scale, higher indicating better status. A score <60 indicates poor pain or function (24–26). Medication use was standard of care; biologic DMARDs were stopped prior to surgery and the last date of use was recorded. Medications withheld for surgery were considered in current use and were typically restarted week 4–6.

Assessment of Flares:

The Rheumatoid arthritis flare questionnaire (RA-FQ), which has been validated in clinical and observational trials and includes questions about severity and duration of flares, self-management, and medication use, was administered at baseline and weekly for six weeks post-operatively. A score of >7 indicates a severe flares (15, 27–29). Flares were identified by self-report as a response to the anchor question “Are you in a flare? Yes/No” and confirmed by concordance between patient report and MD assessment of concurrent RAPID3, and the patient reported painful and swollen joint region score modified from the RADAI questionnaires on a homunculus depicting 40 joints. All aggregate data including baseline data were assessed for each patient independently by 2 investigators (SG, VB, or BM) after the 6-week visit to assign an MD assessment of inflammatory worsening (flare -yes or no), with high concordance between investigators (11).

Data analysis:

Baseline characteristics were summarized as frequencies and percentages for categorical variables; mean and standard deviation for continuous variables. Categorical variables were compared using Chi-squared or Fisher’s exact test. Continuous variables were compared using two-sample t-test or Wilcoxon rank-sum test. One-year outcomes (HOOS/KOOS pain, function, and the change in pain and function) were compared based on flare status at the time of surgery. Since these data were not normally distributed by Shapiro-Wilk test, median and interquartile range were used for the summary; and they were compared using Wilcoxon rank-sum test. Spearman correlation coefficients (SCC) were calculated to evaluate the associations between disease activity and one-year HOOS/KOOS pain and function outcomes since all these data were not normally distributed by Shapiro-Wilk test, with an absolute value >0.7 indicating a strong correlation. Univariate and multivariate regression analysis was performed to explore the relationship between predictors and disease activities at one year. Linear regression analyses were performed; RAPID3, DAS28 and CDAI were transformed to categorical variables: high/moderate vs. low/remission (reference group), and logistic regression analyses were performed. Predictors that were found statistically significant (p<0.05) in the univariate analysis were included in the multivariate analysis; flare status was forced into the model as the variable of interest The final models for multivariate analysis were chosen using backward selection method.

Results

We included 122 patients in the final analysis; 315 initially gave consent; 129 were lost to follow up and 64 had not yet reached the 1 year exam at the time of this analysis (Figure 1). There was no statistically significant difference in the baseline characteristics and disease activities of patients who were included (n=122) and were not included (n=193) in the study, except age, race, RADAI and MD global scores. Specifically, the patients who were included in this study were younger (61.50±11.35 vs. 64.47±11.09, p=0.01), more likely to be white 80 vs. 56% p=0.001), with relatively higher RADAI (9.82±7.99 vs. 5.98±7.47, p=0.001) and with relatively lower MD Global score (4.23±1.99 vs. 4.68±1.72, p=0.04) (Appendix Table 2). Of the 122 patients included in the final analysis; 104 (85%) were female, 98 (80%) were white, and 56 (46%) were undergoing THA. There was no statistically significant difference in the baseline characteristics and disease activity between the patients undergoing TKA vs. THA (Appendix Table 3), except RA criteria and biologic medication treatment. Specifically, TKA patients were more likely to meet ACR RA criteria than THA patients (89% vs. 72%, p=0.04); TKA patients were more likely to be treated with biologic medications than THA patients (59% vs. 39%, p=0.03).

Figure 1.

Recruitment Flow Chart

Of 117 patients who reported their flare status, 68 (58.1%) of patients flared within six weeks of surgery, of whom 37 (54.41%) had flare at baseline. More patients with flares were on biologics and glucocorticoids; methotrexate use was no different between groups. Those who reported flares had higher baseline disease activity including DAS28 ESR (non-flare 3.07 ± 1.04 vs. flare 3.88 ± 1.29, p=0.0002) and statistically and clinically significantly worse scores for HOOS/KOOS pain and function than those who did not flare(Table 1). The absolute change in HOOS/KOOS pain and HOOS/KOOS function from baseline to 1 year was no different in those that flared; both achieved clinically significant improvements (Table 2). Pain scores improved >40 points (non flares 42.5+/- 20 vs flares 47.4 +/- 28.8, p=0.25) and function scores improved ≥ 35 points (non-flares 35.3 +/- 33.8 vs flares 39.05 +/- 26.35, p=0.09), with no significant difference between groups. However, despite significant improvement, HOOS/KOOS pain scores at 1 year were clinically and statistically significantly worse in flarers (non-flare 97.20±13.90 vs. flare 87.5±20.0, p=0.02). HOOS/KOOS function at 1 year was not significantly different between the groups (non-flare 93.35±20.60 vs. flare 86.80±20.60, p=0.07). Flare severity was no different between those with poor HOOS/KOOS pain or function at 1 year compared to those with either no flare or less severe flares.

Table 1:

Baseline Characteristics

Variable, level		Non-flares (N=49) Mean ±SD or N (%)	Flares (N=68) Mean ± SD or N (%)	p-value
Age		63.44 ± 10.21	60.32 ± 10.76	0.12
Sex, Female		41 (83.67)	59 (86.76)	0.64
Race, White		46 (93.88)	48 (70.59)	0.01
BMI		27.69 ± 6.26	29.18 ± 7.18	0.25
RA Criteria,	Criteria 1987	7 (14.58)	9 (13.85)	0.37
	Criteria 2010	15 (31.25)	14 (21.54)
Methotrexate, Yes		26 (53.06)	35 (52.24)	0.93
Steroids, Yes		17 (34.69)	28 (41.79)	0.44
Biologic, Yes		22 (44.90)	36 (53.73)	0.35
RF result interpretation, Positive		20 (41.67)	26 (38.24)	0.93
Anti-CCP interpretation, Positive		35 (71.43)	46 (67.65)	0.94
MDHAQ at baseline		3.31 ± 1.62	4.11 ± 1.35	0.02
RAPID3 at baseline		12.77 ± 6.06	17.07 ± 4.09	0.0002
RADAI at baseline		6.96 ± 5.85	12.26 ± 8.53	0.0003
MDSJC at baseline		3.57 ± 3.61	5.43 ± 6.02	0.17
MDTJC at baseline		2.77 ± 4.44	3.78 ± 5.23	0.18
DAS28-ESR at baseline		3.07 ± 1.04	3.88 ± 1.29	0.002
HOOS/KOOS pain at baseline		46.56 ± 20.43	33.22 ± 16.50	0.004
HOOS/KOOS function at baseline		51.98 ± 22.88	39.73 ± 16.87	0.02

SD = standard deviation; BMI = body mass index; RF = rheumatoid factor; CCP = cyclic citrullinated peptide; MDHAQ = multi-dimensional health assessment questionnaire; RAPID3 = Routine Assessment of Patient Index Data; RADAI = RA Disease Activity Index; MDSJC = MD swollen joint count; MDTJC = MD tender joint count; DAS28-ESR = DAS28-erythrocyte sedimentation rate; HOOS/KOOS = hip disability/knee injury osteoarthritis outcome score

Table 2:

Comparison of one-year outcome based on flares at the time of the surgery

Variable	Non-flares (n=49) (Median ± IQR)	Flares (n=68) (Median ± IQR)	p-value
HOOS/KOOS pain at 1 year	97.20±13.90	87.5±20.00	0.02
HOOS/KOOS function at 1 year	93.35±20.60	86.80±20.60	0.071
Change in HOOS/KOOS pain 1 year from baseline	42.50±20.0	47.40±28.80	0.25
Change in HOOS/KOOS function 1 year from baseline	35.30±33.80	39.05±26.35	0.088
HAQ at 1 year	1.70±2.30	2.30±2.00	0.04

IQR = interquartile range; HOOS/KOOS = hip disability/knee injury osteoarthritis outcome score; HAQ = health assessment questionnaire

We next sought to determine whether baseline RA disease activity or other clinical features are associated with HOOS/KOOS pain and function at 1 year. We used Spearman’s correlations and revealed associations between baseline disease activity and HOOS/KOOS pain and function at 1 year (Appendix Table 4). DAS28 at baseline and HOOS/KOOS function at 1 year were significantly negatively correlated (Spearman’s rho=-0.46, p-value<0.0001 indicating that lower disease activity was associated with improved pain and function

We constructed linear regression models to explore the association of baseline disease activity, flares, or other clinical features and HOOS/KOOS pain scores at 1 year. White race, BMI, flare, DAS28 and baseline HOOS/KOOS pain and function all predicted HOOS/KOOS pain at 1 year in the univariate models (Table 3). In the multivariate model, DAS28 and BMI remained statistically significant after adjusting for the other clinical features (Table 4). Specifically, each 1 unit increase in DAS28 worsens the pain score at 1 year by 2.41 unit with p=0.02, and each 1 unit increase in BMI worsens the pain score at 1 year by 0.54 (p=0.004). While BMI is a consistent and statistically significant finding, this BMI association translates to a requirement of approximately 20 units of difference in BMI to increase the pain score by 10 points, a minimally clinically significant change. Notably, there was no statistically significant difference in the risk of worse pain at 1 year between those who reported post-operative flares compared to those who did not (p=0.22). We also examined clinical predictors of function one year after arthroplasty. In a univariate linear regression model built to explore predictors of HOOS/KOOS function one year after surgery, baseline DAS28, MDHAQ, RAPID3, BMI, and HOOS/KOOS function were all significantly associated with worse postoperative HOOS/KOOS function (Table 5). In multivariate analysis, DAS28 and BMI remained statistically significantly associated with HOOS/KOOS function one year after surgery (Table 6). Specifically, increasing DAS28 by 2.1 units will decrease the function score at 1 year by >10 points (Estimate −4.82 (Standard error 1.17. p=0.0001). Each 1 unit increase in BMI will worsen the function score at 1 year by 0.46 (p=0.03), a clinically insignificant change without a massive difference in BMI. There was no statistically significant difference in the risk of poor function at 1 year between those reporting post-operative flares compared to those who did not (p=0.98).

Table 3:

Univariate Analysis – Outcome Variable: HOOS/KOOS pain score at 1 year, Linear Regression; n=115

Univariate variable, level		Estimate (standard error)	p-value
Age		0.26 (0.11)	0.02
Sex,	Female vs. Male	−6.79 (3.49)	0.06
Race,	Black vs. White	−11.96 (4.35)	0.007
	Other vs. White	−6.16 (4.57)	0.18
Ethnicity, Hispanic or Latino vs Not Hispanic nor Latino		−10.35 (5.15)	0.04
BMI		−0.47 (0.18)	0.01
RA Criteria,	criteria 1987 vs Meets BOTH criteria	1.26 (4.02)	0.76
	criteria 2010 vs Meets BOTH criteria	−1.49 (3.17)	0.64
	does not meet criteria vs. Meets BOTH criteria	4.92 (3.62)	0.18
Steroids, No vs. Yes		−1.56 (2.58)	0.55
Biologic, No vs. Yes		0.73 (2.54)	0.77
Flare, No flare vs Yes flare		6.16 (2.58)	0.02
DAS28-ESR at baseline		−2.94 (1.02)	0.005
CDAI at baseline		−0.19 (0.12)	0.12
MDHAQ at baseline		−1.51 (0.81)	0.07
RAPID3 at baseline		−0.43 (0.23)	0.07
RADAI at baseline		−0.29 (0.15)	0.06
MDTJC at baseline		−0.50 (0.27)	0.07
HOOS/KOOS pain at baseline		0.19 (0.08)	0.02
HOOS/KOOS function at baseline		0.15 (0.06)	0.02

HOOS/KOOS = hip disability/knee injury osteoarthritis outcome score; BMI = body mass index; RA = rheumatoid arthritis; DAS28-ESR = disease activity score 28-erythrocyte sedimentation rate; CDAI = clinical disease activity index; MDHAQ = multi-dimensional health assessment questionnaire; RAPID3 = routine assessment of patient index data; RADAI = RA disease activity index; MDTJC = MD tender joint count

Table 4:

Multivariate analysis – Outcome variable: HOOS/KOOS pain score at 1 year, Linear Regression; n=115

Multivariate variable	Level	Estimate (standard error)	p-value
DAS28-ESR		−2.41 (1.05)	0.02
BMI		−0.54 (0.18)	0.004
Flare	No flare vs Yes flare	3.43 (2.76)	0.22

HOOS/KOOS = hip disability/knee injury osteoarthritis outcome score; DAS28-ESR = disease activity index 28-erythrocyte sedimentation rate; BMI = body mass index

Table 5:

Univariate Analysis – Outcome variable: HOOS/KOOS function score at 1 year, Linear Regression; n=118

Univariate variable, Level		Estimate (standard error)	p-value
Age		0.12 (0.14)	0.38
Sex,	Female vs. Male	−8.01 (4.41)	0.07
Race,	Black vs. White	−6.05 (5.80)	0.30
	Other vs. White	−2.33 (6.08)	0.70
Ethnicity, Hispanic or Latino vs Not Hispanic nor Latino		−7.82 (6.73)	0.25
BMI		−0.65 (0.23)	0.01
RA Criteria,	criteria 1987 vs Meets BOTH criteria	4.02 (4.54)	0.38
	criteria 2010 vs Meets BOTH criteria	−1.84 (3.66)	0.62
	Does not meet criteria vs. Meets BOTH criteria	4.47 (4.11)	0.28
Steroids, No vs. Yes		2.36 (3.30)	0.48
Biologic, No vs. Yes		−1.32 (3.23)	0.68
Flare, No flare vs Yes flare		5.80(3.32)	0.08
DAS28-ESR at baseline		−4.96 (1.11)	0.0001
CDAI at baseline		−0.27 (0.16)	0.08
MDHAQ at baseline		−3.74 (1.00)	0.0003
RAPID3 at baseline		−0.84 (0.29)	0.004
RADAI at baseline		−0.25 (0.20)	0.21
MDJTC at baseline		−0.72 (0.34)	0.035
HOOS/KOOS pain at baseline		0.16 (0.09)	0.08
HOOS/KOOS function at baseline		0.25 (0.08)	0.003

HOOS/KOOS = hip disability/knee injury osteoarthritis outcome score; BMI = hip disability/knee injury osteoarthritis outcome score; RA = rheumatoid arthritis; DAS28-ESR = disease activity score 28-erythrocyte sedimentation rate; CDAI = clinical disease activity index; MDHAQ = multi-dimensional health assessment questionnaire; RAPID3 = routine assessment of patient index data; RADAI = RA disease activity index; MDTJC = MD tender joint count

Table 6:

Multivariate analysis – Outcome Variable: HOOS/KOOS Function score at 1 year, Linear Regression; n=118

Multivariate variable	Level	Estimate (standard error)	p-value
DAS28-ESR		−4.82 (1.17)	0.0001
BMI		−0.46 (0.21)	0.03
Flare	No flare vs Yes flare	0.087 (3.07)	0.98

HOOS/KOOS = hip disability/knee injury osteoarthritis outcome score; DAS28-ESR = disease activity score 28-erythrocyte sedimentation rate; BMI = body mass index

Finally, we determined if baseline clinical features could predict disease activity scores (RAPID3 and RADAI) one year after surgery. Clinical features included in this analysis were baseline patient characteristics and disease activity, in two separate models, for RAPID3, baseline scores for RAPID3, RADAI, DAS 28, HOOS/KOOS pain and function, biologic therapy, and positive flare were significant in the univariate model (Appendix Table 5) and were included in the multivariable logistic regression model (Appendix Table 6). Only HOOS/KOOS function at baseline predicted RAPID3 at 1 year (OR 0.96 (95% CI 0.94, 0.99). For RADAI, similarly, race, BMI, flare status, baseline pain, function, RADAI, MDTJC, DAS28_ESR, and CDAI were significant in the univariate model, (Appendix Table 7), and included in the final model. HOOS/KOOS pain (Estimate=-0.16 (SE 0.07) p=0.02), RADAI (Estimate 0.35 (SE 0.13) p=0.02) and CDAI (Estimate 0.27 (SE 0.10) p=0.01) at baseline were statistically significantly associated with RADAI at 1 year (Appendix Table 8). Overall, those with more active RA or whose pain or function was worse at baseline were significantly more likely to have higher disease activity at 1 year, but the effect sizes were small.

Discussion

We here report the relationship of active or worsening inflammatory disease in patients with RA at the time of THA and TKA or in the post-operative period and analyze the effect on HOOS/KOOS pain and function outcomes one year later. We found higher DAS28 scores and BMI increase the risk of worse function and higher DAS28 scores additionally increase the risk of worse pain one year after THA and TKA. Multiple prior studies have demonstrated that patients with RA have less improvement in pain and function after THA and TKA than patients with OA, but whether RA disease activity was a risk factor for worse outcomes was not known(30, 31). Given the majority of patients are obese and have DAS28 greater than 3.2 at the time of THA/TKA, these may be important modifiable risk factors (32). Although patients who flared had worse pain scores at one year, perioperative flares were not an independent risk factor for worse outcomes independent of baseline disease activity, in the multivariable model, and those who flared improved as much as those who did not flare. As flares did not increase the risk of worse outcomes perioperative medication management decisions may be undertaken without concern for long-term impact on arthroplasty outcome linked to flares. As higher disease activity at baseline was the strongest predictor of pain and function at 1 year, one can speculate that a flare is less significant than sustained high disease activity.

Previous studies have reported on THA/TKA outcomes for patients with RA, and determined that patients with RA are at risk for worse pain and function when compared to patients with OA. While important confounders such as age, sex, and comorbidities have been considered, disease activity and RA status have not been included in analysis of arthroplasty outcomes such as HOOS and KOOS(30, 33, 34). In a study that included DAS28 measurements pre and post op, no improvement in function measured as HAQ was observed, but this finding may reflect the recognized floor effect for the HAQ (22, 35). Negative correlations of post-operative disease activity and post-operative function have been reported, but the impact of pre-op disease activity, a potentially modifiable risk factor, has not been described(35–37). In our cohort, while we demonstrated that higher disease activity and worse HOOS/KOOS pain and function scores at the time of arthroplasty predicted higher disease activity at 1 year, we also identified higher DAS28 as an independent risk factor for worse outcomes. We were interested in the overall effect of THA/TKA on disease activity at 1 year. While HOOS/KOOS function at baseline predicted RAPID3 at 1 year, and HOOS/KOOS pain, RADAI and CDAI predicted RADAI at 1 year the effect sizes were small.

This study has certain limitations. There was significant loss in follow up, predominantly among older, non-white patients with slightly more active disease, but other baseline characteristics were not significantly different. In addition, as our primary outcome was the relationship between disease activity and HOOS/KOOS at 1 year, this is unlikely to have biased our results. While DAS28 at baseline and BMI predicted pain and function at one year, the effect size, particularly for BMI, was small. We pooled THA and TKA results and although DAS28 assessments do not include the hip, no significant difference is reported in DAS28 disease activity post THA compared to post TKA, suggesting that this is unlikely to be a significant source of bias (34). Our flare designation was based on patient report in the RA-FQ, as it was not feasible to perform weekly physical exams. Instead, two rheumatologists assessed concurrently administered disease activity questionnaires (RAPID3 and RADAI) as well as baseline clinical data to validate patient reports of flares. While this data was collected in a single tertiary referral center, and may not be generalizable, it is unlikely that the relationship of RA disease activity to 1 year outcomes would be affected by the single center design. Additionally, HOOS/KOOS measure lower extremity pain and function, so those with active RA might have lower scores due to involvement of other lower extremity joints, and the low baseline pain and function scores may have reflected higher disease activity. Moreover, while HOOS/KOOS are joint specific, RA is typically symmetric; however, these instruments, which contain the Western Ontario and McMaster Universities Arthritis Index (WOMAC), have been widely used in RA (8). Finally, as with any observational study, our results may be skewed by unmeasured confounders.

There are strengths of this study that warrant mention, including the prospective study design and the combined use of both patient reported outcomes and measures including MD joint counts and inflammatory markers. As not all patients return for a 1 year evaluation, we administered a validated PROM, the HOOS/KOOS, that is widely used to determine THA and TKA outcomes at 1 year. Post-operative flares were defined by prospectively administering a weekly flare questionnaire for 6 weeks after surgery, reducing the risk of recall bias. We additionally administered the RAPID3 and RADAI joint symptom questionnaires and self-reported specific joint counts excluding the operative joint to reduce conflation of post-operative pain with RA disease activity.

In summary, we found that increased DAS28 and BMI, but not perioperative flares, increase the risk for worse pain and function one year after surgery. Given both of these features are modifiable, and the majority of patients with RA have moderate to high disease activity at the time of arthroplasty, future efforts are needed to determine whether pre-arthroplasty medical management should include efforts to control disease activity and weight, and whether these interventions might improve long term pain and function.

Significance and Innovation.

Innovation

• The relationship of peri-operative disease activity and flares of RA with pain and function scores one year after THA/TKA is unknown.

Significance

• Understanding the reason for poor pain and function scores of RA patients compared to those with OA can ultimately help optimize post-operative care for RA patients.

• Those with high disease activity at the time of THA/TKA experience improvements in pain and function, but less than those with well-controlled RA.