Targeted and biologic therapies and risk of total knee or hip replacement in axial spondyloarthritis and psoriatic arthritis

Abstract

Background:

Axial spondyloarthritis (axSpA) and psoriatic arthritis (PsA) are chronic inflammatory diseases that often cause joint damage, potentially leading to joint replacement surgery. We assessed whether Janus kinase (JAK) inhibitors reduce the risk of total knee or hip replacement compared to nonsteroidal anti-inflammatory drugs (NSAIDs).

Methods:

Using the Merative^™ MarketScan^® Commercial Database, we conducted a nested case-control study of adults aged 18–65 years with axSpA and/or PsA from October 2015 to December 2021. Medication exposure was categorized hierarchically using pharmacy and procedure claims, including JAK inhibitors, non-tumor necrosis factor inhibitor biologics (non-TNFi biologics), TNF inhibitors (TNFi), DMARDs, NSAIDs (referent), and none. Logistic regression with confounder adjustment assessed associations between medication class and joint replacement risk.

Results:

Among 8,855 eligible adults, 1,771 cases of joint replacement were identified. JAK inhibitor use was not significantly associated with reduced odds of joint replacement compared to NSAIDs (odds ratio [OR] 0.67, 95% confidence interval [CI] 0.41–1.08). Non-TNFi biologic users (OR 0.66, 95% CI 0.53–0.82), TNFi users (OR 0.63, 95% CI 0.52–0.76), and DMARD users (OR 0.65, 95% CI 0.53–0.80) had lower odds of joint replacement than NSAID users.

Conclusion:

We did not find conclusive evidence that relative to NSAIDs, JAK inhibitors prevent end-stage arthritis requiring surgery in axSpA and PsA, however risk was reduced with use of non-TNFi biologics, TNFi, or DMARDs. Longer term data are needed to understand the optimal utilization of JAK inhibitors in preventing end-stage arthritis in these conditions.

Introduction

Axial spondyloarthritis (axSpA) and psoriatic arthritis (PsA) are chronic inflammatory conditions characterized by diverse clinical presentations that include inflammatory arthritis. Up to a third of individuals with axSpA and most individuals with PsA have peripheral joint disease, with some developing secondary osteoarthritis (OA) of the peripheral joints that results in end-stage arthritis requiring arthroplasty [1, 2]. Among those with axSpA, 49% have large lower extremity joint involvement (hip/knee), while for those with PsA, hip involvement is around 21% [3]. The prevalence of joint replacement, in particular total knee replacement (TKR) and total hip replacement (THR), has not been well studied but is thought to be elevated in these patients in comparison to the general population [4, 5].

Current guidelines for axSpA management advocate for non-steroidal anti-inflammatory drugs (NSAIDs) as first-line pharmacotherapy, reserving biologics such as tumor necrosis factor inhibitors (TNFi) and IL-17 inhibitors as second- or third-line treatment for individuals experiencing sustained high disease activity despite other treatment [6]. In PsA, the latest guidelines recommend initiating treatment with conventional synthetic disease-modifying antirheumatic drugs (hereafter DMARDs) such as methotrexate, while biologics, including TNFi and IL-17 inhibitors, are considered an alternative when there is inadequate response to DMARDs [7]. Janus kinase inhibitors (JAK inhibitors) are targeted synthetic disease-modifying antirheumatic drugs with trial data demonstrating efficacy in axSpA and PsA [8, 9]. Current guidelines recommend JAK inhibitors’ use after TNFi and other targeted biologics, partly due to potential toxicity (e.g. thromboembolic events and malignancy), particularly among higher risk individuals including those of older age or with cardiovascular comorbidity [6, 7]. However, it remains unclear whether JAK inhibitors may provide additional benefit over current first-line NSAIDs therapy or other medications with different mechanisms of action in postponing or preventing end-stage large joint arthritis in patients with axSpA or PsA, particularly given the paucity of head-to-head trials comparing these therapies.

While there are hypothetical effects of Janus kinase inhibition on bone metabolism, there are limited data regarding the potential of JAK inhibitors on preventing structural bone and joint damage and reducing the risk of joint replacement in individuals with axSpA and PsA [10]. Therefore, the aim of our study was to investigate whether, among subjects with axSpA and PsA, the use of JAK inhibitors is associated with a reduced risk of joint replacement surgery (TKR or THR) as a surrogate for end-stage arthritis compared to use of NSAIDs. Assessing the impact of this medication class can provide additional evidence to support current guidelines and assist in the use of a personalized treatment strategy for those with axSpA or PsA.

Materials and methods

Data Source

We used data from the Merative^™ MarketScan^® Commercial Database, which consists of medical and drug data from employers and health plans. It contains data for over 250 million individuals, including employees, their spouses, and dependents who are covered by employer-sponsored private health insurance in the US [11].

Ethical approval

As this study involved the analysis of pre-existing, de-identified data, it was deemed exempt from Institutional Review Board review. (BUMC IRB H-42900).

Study sample

Insurance claims were utilized to identify adults aged between 18 and 65 years with axSpA and PsA between whose data was in Marketscan between October 2015 and December 2021. AxSpA (including both radiographic and nonradiographic axSpA) or PsA were operationally defined as ≥1 inpatient claim with an ICD-9 or ICD-10 diagnosis code or ≥2 outpatient claims ≥7 days apart. Such algorithms for identification of axSpA have previously been shown to have high positive predictive value (88-100%) [12, 13]. If a subject had both axSpA and PsA diagnoses, they were classified as axSpA, and the disease date was set to that of the first claim. Included individuals needed to have a year of continuous enrollment prior to their event date.

Study design

We conducted a nested case-control study. Study entry date was the date of the first inpatient claim or the date of the first outpatient claim for axSpA or PsA. Included individuals were followed until the outcome occurred, age 65 (age at which patients became eligible to Medicare coverage), coverage end date, or study end date (December 31, 2021).

Outcome of interest

The outcome was THR or TKR defined using procedure codes in any position. The assumption was made that subjects underwent THR/TKR due to end-stage arthritis secondary to axSpA and/or PsA. Joint replacement has been used as a proxy for end-stage arthritis in these conditions in other studies [14, 15]. Cases were designated as individuals with axSpA or PsA who underwent THR or TKR, while controls were defined as individuals with axSpA and/or PsA who did not undergo such procedures prior to the event date. The event date for cases was set as the date of the first THR/TKR during the study period. For controls, the event date was the matching date of the cases.

Exposure of interest

We evaluated axSpA and/or PsA treatment during the 6-month period preceding the event date (exposure assessment period). Individuals who had been exposed to multiple medication classes were grouped hierarchically: JAK inhibitors > non-TNFi biologics> TNFi > DMARDs > NSAIDs > no medication. This classification scheme was developed to be consistent with clinical practice guidelines and to mimic clinical practice patterns [6,7,16]. Medications are included in Supplemental Table 1. The exposure date was set to the first prescription of the exposure drug within the 6-month exposure assessment period (Figure 1). For individuals with no relevant medication exposures, there was no exposure date. TNFi use was determined through a combination of pharmacy claims and Current Procedural Terminology (CPT) codes on medical claims for infusions, while all other medications were assessed solely through pharmacy claims due to administration orally or by subcutaneous injection. NSAID use was defined as the reference group, as this was the most commonly used medication class for the study population during the time period studied. In guidelines for the management of axSpA, NSAIDs are recommended as first-line pharmacologic treatment, and for PsA they are recommended for relief of musculoskeletal signs and symptoms

Figure 1. Study design (nested case-control).

Abbreviations: CE, continuous enrollment; axSpA, axial spondyloarthritis; PsA, psoriatic arthritis; THR, total hip replacement; TKR, total knee replacement; TNFi, tumor necrosis factor inhibitor; JAKibs Janus kinase inhibitors; DMARDs, disease modifying anti-rheumatic drugs

Event date: joint replacement date for cases, and random date from the corresponding case’s joint replacement year for controls

Disease date: can be any time prior to exposure assessment period

Continuous Enrollment (CE) Period: must include at least 1 year prior to event date

Exposure Assessment Period: 6 months prior to event date

Covariate Assessment Period: 6 months prior to exposure assessment period

Covariates

Potential confounders for the relationship between axSpA/PsA medication class and joint replacement were identified a priori. For individuals with relevant medication exposure, the covariate assessment period comprised the 6 months prior to the medication exposure date (Figure 1). For subjects without a relevant medication exposure, the covariate assessment period comprised the 6 months prior to the index date. Demographic confounders included age and sex. Other confounders included body mass index (BMI) and comorbidities such as baseline hip and/or knee OA, history of chronic kidney disease, cardiovascular disease, diabetes, hypertension, alcohol use, and tobacco use. Direct measures of PsA or axSpA disease severity, disease activity, or disease duration are not available in MarketScan. Confounders that were surrogates of axSpA or PsA disease activity included erythrocyte sedimentation rate (ESR)/C-reactive protein (CRP) laboratory orders, and outpatient visits (rheumatology and primary care).

With the exception of age and BMI, covariates were assessed within 6 months prior to the exposure date in the covariate assessment period (Figure 1). Age was assessed on the event date. BMI was based on categorical variables defined as the most recent documentation within the 6 months prior to the exposure assessment period. Specific ICD-10 codes were used for underweight, normal weight, overweight/obese. Since secondary OA may be a result of structural damage from axSpA or PsA, we assessed the presence of baseline knee or hip OA prior to the date of axSpA or PsA diagnosis using ICD codes for OA.

Statistical analyses

For each THR or TKR case, 4 controls were included, based on the following criteria: control subjects’ disease date did precede the exposure assessment period, control subjects were actively enrolled on the event date of cases, and they had at least one year of continuous enrollment prior to the event date. Subjects were included using risk-set sampling, allowing subjects to be included in the same sample multiple times with different event dates, only once as a case and multiple times as a control. Additionally, it was possible for a control subject to experience the outcome after their event date.

We did not match on other potential confounders such as age because matching in a case-control design does not inherently control for confounding by the matching variables and can introduce new confounding by these variables, even if none existed in the source population [17]. Moreover, matching in case control studies can limit the generalizability of findings by restricting inferences to the matched sets.

We examined the odds of TKR/THR comparing axSpA/PsA medication class versus NSAID use using unconditional logistic regression. Models were adjusted for potential confounders noted above except for baseline OA (Model 1) and all potential confounders including baseline OA (Model 2).

To assess the robustness of our results, we performed the following sensitivity analyses: stratified by axSpA versus PsA, restricted to TKR only, and restricted to individuals without prior history of joint replacement.

Results

In the main nested case-control analysis, we included 8,855 adults, of which 2,488 had axSpA and 6,367 had PsA, comprising 1,771 cases of TKR/THR with 7,084 matched controls (Figure 2). Among cases, the mean age was 56.4 (± 6.4) years, with 56.6% being female, whereas for controls, the mean age was 49.5 (± 10.3) years, with 52.2% being female (Table 1). Compared to controls, cases had a numerically higher frequency of comorbidities including chronic kidney disease (3.8% vs 2.5%), cardiovascular diseases (12.7% vs 7.5%), diabetes (15.1% vs 12.7%), hypertension (46.2% vs 30.7%), alcohol use disorder (1.4% vs 0.7%), and tobacco use (5.1% vs 3.6%). Among those with reported BMI, the case group had a higher prevalence of overweight/obese subjects (27.8% vs 21.1%) (Table 1).

Figure 2. CONSORT flow diagram of study inclusion.

Abbreviations: axSpA: Axial Spondyloarthritis; PsA: Psoriatic arthritis; THR: total hip replacement; TKR: total knee replacement

Table 1.

Baseline demographics and clinical characteristics of the case-control sample of individuals with axial spondyloarthritis and psoriatic arthritis

		Cases (THR/TKR)	Controls
Subjects, n		1771	7084
Age, years, mean ± SD		56.4 ± 6.4	49.5 ± 10.3
Female, n (%)		1003 (56.6%)	3701 (52.2%)
Disease, n (%)	AxSpA	462 (26.1%)	2026 (28.6%)
	PsA	1309 (73.9%)	5058 (71.4%)
Chronic kidney disease, n (%)		68 (3.8%)	176 (2.5%)
Cardiovascular disease, n (%)		225 (12.7%)	530 (7.5%)
Diabetes, n (%)		267 (15.1%)	903 (12.7%)
Hip/Knee osteoarthritis, n (%)		594 (33.5%)	566 (8.0%)
Hypertension, n (%)		818 (46.2%)	2177 (30.7%)
Alcohol use disorder ~ , n (%)		25 (1.4%)	53 (0.7%)
Body mass index	Underweight	5 (0.3%)	23 (0.3%)
	Normal	16 (0.9%)	132 (1.9%)
	Overweight/Obese	492 (27.8%)	1493 (21.1%)
	Missing	1258 (71.0%)	5436 (76.7%)
Tobacco use ^ , n (%)		90 (5.1%)	257 (3.6%)
ESR/CRP lab ordered, n (%)		922 (52.1%)	3224 (45.5%)
Rheumatology visit # , n (%)		924 (52.2%)	3429 (48.4%)
Primary care visit # , n (%)		1358 (76.7%)	4683 (66.1%)

Abbreviations: ESR: erythrocyte sedimentation rate; CRP: C-reactive protein; axSpA, axial spondyloarthritis; PsA, psoriatic arthritis; THR/TKR: total hip replacement/total knee replacement

^{^}Alcohol use disorder and tobacco use (including smoking) as defined by ICD codes in the Chronic Conditions Data Warehouse

^#Primary care and rheumatology visits were defined as the number of individuals with at least one visit that occurred any time the 6-month period prior to the 6-month exposure assessment period.

Among THR/TKR cases, use of JAK inhibitors was 1.5% vs 1.2% in controls, non-TNFi biologics was 13.5% vs 13%, TNFi was 35.8% vs 37%, DMARDs was 18.4% vs 14.1%, NSAIDs was 13.9% vs 8.6%, and no exposure to any of the medications was 16.9% vs 26.1%. Patients who were either not receiving any treatment or were exclusively on NSAIDs had fewer visits to the rheumatologist compared to those on other classes of treatment (Supplementary Table 2). Those not on treatment also had a lower frequency of laboratory investigations for markers of inflammation, such as ESR and CRP.

In multivariable models, when adjusted for potential confounders except for baseline OA, patients receiving JAK inhibitors had 33% lower odds of undergoing joint replacement compared to those receiving NSAIDs, though this was not statistically significant (OR 0.67, 95% CI 0.41–1.08). Several medication classes were associated with a significantly decreased odds of TKR/THR including non-TNFi biologics with 34% reduced odds (OR 0.66, 95% CI 0.53–0.82), TNFi with 37% reduced odds (OR 0.63, 95% CI 0.52–0.76), and DMARDs with 35% decreased odds (OR 0.65, 95% CI 0.53–0.80) compared to NSAIDs users. Patients who did not receive any treatment had 59% lower odds (OR 0.41, 95% CI 0.34–0.51) of undergoing joint replacement. These values remained consistent after further adjustment for baseline hip and/or knee OA (Figure 3; Supplementary Table 3).

Figure 3. Results of multivariate analysis for the association of medication class and total hip or knee replacement.

JAKibs include Baricitinib, Tofacitinib, Upadacitinib.

TNFis include etanercept, adalimumab, golimumab, certolizumab and infliximab.

The results of sensitivity analyses indicated consistent results when stratified by disease (axSpA vs PsA) (Supplementary Table 4), restriction to TKR outcomes only (Supplementary Tables 5-6), and restriction to individuals with no history of prior total knee/hip OA (Supplementary Tables 7-8).

Discussion

Amid ongoing discussion regarding the influence of advanced therapies on peripheral arthritis in axSpA and PsA patients, our study investigated the potential effects of JAK inhibitors and other targeted therapies in averting end-stage arthritis requiring THR/TKR. While the effect of JAKibs did not achieve statistical significance, we observed protective effects (34-37% reduction in odds) with non-TNFi biologics, TNFi, and DMARDs on end-stage arthritis.

JAK inhibitors are recommended in clinical practice guidelines for individuals with axSpA who have shown inadequate response to previous medication regimens [18]. While there is limited literature on the effects of JAK inhibitors usage on joint replacement, clinical trials have outlined a favorable impact of JAK inhibitors in ameliorating inflammation and limiting disease activity in axSpA, as evidenced by reduction in MRI inflammation scores and disease activity scores [19]. Moreover, in populations with PsA, JAK inhibitors have shown efficacy in improving disease activity (measured by the ACR20) and physical function [20], in one study demonstrating superiority to adalimumab with higher doses [21]. Given the evidence that JAK inhibitors improve disease activity in axSpA and PsA, it was plausible that this class of medications could potentially reduce the rates of THR/TKR.

Limited data exist on progression to end-stage arthritis in peripheral joints for axSpA or PsA when comparing JAK inhibitors to NSAIDs or other classes of medications. While the available data on medication use in axSpA and PsA provide valuable insights into control of symptoms and inflammation, and on radiographic progression, they have not explicitly examined the relation between medication use and the clinically important outcome of end-stage arthritis [22, 23]. Our findings indicate that JAK inhibitors were not associated with a statistically significant reduction in the risk of THR/TKR in comparison to NSAIDs. On one hand, this may be explained by the relatively limited data and low number of THR/TKR events leading to limited precision and thus wider confidence intervals. Another explanation for the absence of a clear protective effect is that individuals prescribed advanced therapies such as JAK inhibitors have often accumulated significant joint damage from uncontrolled inflammation, making them at higher risk of joint destruction and irreversible end-stage arthritis. In treatment guidelines [6,7,24], JAK inhibitors are often prescribed as third- or fourth-line therapy, and therefore, JAK inhibitor users may have had more longstanding arthritis (i.e., longer disease duration) and/or a more severe form of arthritis characterized by lack of response to other classes of medications, potentially heightening the likelihood of end-stage joint involvement.

Another consideration is that the mean age of our population is higher than a population with newly diagnosed axSpA. Older patients are likely to have more advanced joint damage from their inflammatory arthritis with longer disease duration, and OA is more prevalent with age [25]. However, our results remained consistent after adjusting for OA, which was assessed at baseline (prior to axSpA or PsA diagnosis).

Our findings revealed a 34-37% decreased risk of undergoing TKR/THR users of non-TNFi biologics, TNFi, and DMARDs compared to NSAID users. Previous studies investigating the impact of TNFi on joint replacement in patients with rheumatoid arthritis (RA) have showed inconclusive results overall. However, some evidence suggests a potential benefit in specific subgroups, for instance in patients over 60 years of age, where a reduction in THR rates was observed with TNFi use [26]-[27]. Similarly, the effects of DMARDs on joint replacement outcomes in RA have also been inconclusive across previous studies [28]-[29]. While our findings possibly represent a true protective effect of these medications on arthritis activity in axSpA and PsA, residual confounding remains possible. For example, individuals with cardiovascular risk factors may be preferentially treated with a medication class other than NSAIDs, and may also be less likely to undergo TKR/THR due to surgical risk [30, 31]. It is also possible that NSAIDs (our referent group) could have detrimental effects on the joints increasing the risk of joint replacement. While observational studies in OA have linked prolonged NSAID use to an increased risk of TKR [32,33], literature on the long-term effects of NSAIDs on bone destruction in inflammatory arthritis such as PsA and axSpA is scarce.

Interestingly, patients who did not receive any targeted axSpA/PsA treatment had reduced odds of TKR/THR compared to those on NSAIDs. These individuals potentially had milder arthritis phenotypes that did not require escalation of therapy. However, we acknowledge that this group with no medication exposure could be heterogenous and include those with comorbid conditions and other medication use that was felt to preclude use of immunomodulatory therapies or NSAIDs despite active inflammatory arthritis.

While our study provides valuable insights into the relationship between medication use and joint replacement in patients with axSpA and PsA, several important limitations should be acknowledged. First, a case-control study design was chosen over a cohort design in order to enhance the study’s efficiency to detect a potential association, due to the limited number of cases in each category of medication exposure. Use of a case-control study design did not allow us to account for changes or switches in medication exposure over time.

Next, the observational nature of the study introduces the potential for residual confounding, despite our efforts to control for confounders through matching and adjustment. More than 70% of the BMI data were missing in both the case and control groups, due to its inconsistent collection in administrative claims data. While we also used ICD codes to define BMI classes, our ability to fully account for BMI as a confounder was thus limited. Furthermore, our study likely did not fully address confounding by indication, wherein the observed association between treatment and outcome could be influenced by the underlying indication for treatment rather than the treatment itself. Individuals prescribed JAK inhibitors can be very inherently different from individuals prescribed NSAIDs, and these differences may not be captured in administrative claims data. Disease severity, disease activity, and disease duration are important potential confounders that are not available in insurance claims data. Individuals with worse disease severity or higher disease activity may be more likely to receive JAK inhibitors and also more likely to require joint replacement. To mitigate this residual confounding, we attempted to account for disease severity, activity, and duration via surrogates such as healthcare utilization and measurements of acute phase reactants.

Another consideration is that secondary OA can develop as a consequence of inflammatory arthritis and act an intermediate step in the causal pathway to joint replacement [34]. While studies like ours should account for OA, adjustment for OA captured after the diagnosis of axSpA or PsA may attenuate potential effects. Thus, we assessed for the presence of OA prior to the diagnosis of axSpA or PsA and found that additional adjustments for baseline OA did not measurably change results. Further, we were not able to extract the indications for surgery from our data source, and thus were not able to account for underlying reasons that may lead to joint replacement.

An additional limitation is that our analyses assessed treatment exposure and covariates within a 6-month period before joint replacement surgery, potentially overlooking longer-term treatment. We assumed that the medication used during this period was the primary one affecting the risk of joint replacement during the study period. While it can be argued that this duration might be insufficient to fully study long-term outcomes and that prior treatments may influence the results, clinical trials of JAK inhibitors have shown that significant improvements in disease activity, pain reduction, physical function, and quality of life often begin within a few weeks, and continue to progress through 6 months of therapy. These findings support the notion that a 6-month treatment duration could effectively capture meaningful clinical improvements, reflecting the potential of JAK inhibitors to achieve long-term therapeutic goals in inflammatory arthritis [35, 36].

Finally, the small number of JAK inhibitors users limited the precision of effect estimates (and resulted in wide confidence intervals) and prevented us from evaluating medication exposures more granularly in terms of combination therapy versus monotherapy. Due to small numbers of outcomes, our main analyses combined axSpA and PsA. We acknowledge that the relationship of medication use with end-stage arthritis may differ by disease subtype, which we were unable to study with adequate precision.

Conclusion

In conclusion, our results suggest significant reductions in TKR/THR risk with non-TNFi biologics, TNFi, and DMARDs but not with JAK inhibitors compared to NSAIDs in axSpA and PsA. These findings support the use of non-TNFi biologics, TNFi, and DMARDs both for short term disease control and for long-term joint protection. While JAK inhibitors effectively control disease activity in clinical trials, in real world practice they may either be started too late in the disease process or may not be effective in reducing the risk of end-stage arthritis necessitating surgery. Further research is needed to explore the long-term effects and optimal utilization of JAK inhibitors in preventing end-stage arthritis in these chronic inflammatory conditions.

Key points:

• Axial spondyloarthritis (axSpA) and psoriatic arthritis (PsA) are chronic inflammatory diseases that often cause joint damage, potentially leading to joint replacement surgery.

• In our study, although there was not conclusive evidence that JAK inhibitors prevent end-stage arthritis requiring surgery in axSpA and PsA relative to NSAIDs, we did find that risk was reduced with use of non-TNFi biologics, TNFi, or DMARDs.

• Understanding the impacts of different medication classes, including Janus kinase (JAK) inhibitors and tumor necrosis factor inhibitors (TNFi) relative to nonsteroidal anti-inflammatory drugs (NSAIDs) may guide treatment decisions.