Baseline predictors of disease severity in immune checkpoint inhibitor-induced inflammatory arthritis

Abstract

Objectives

The objective of this study was to determine baseline risk factors for requiring immunosuppression and having persistent arthritis in patients with immune checkpoint inhibitor–induced inflammatory arthritis (ICI–inflammatory arthritis).

Methods

Participants were adults with rheumatologist diagnosed ICI–inflammatory arthritis. The primary outcome was requirement of conventional synthetic (cs) or biologic (b) DMARDs; other outcomes were persistence of inflammatory arthritis >6 months after ICI cessation and requirement of CSs. Logistic regression models evaluated associations between clinical features and primary and secondary outcomes, with adjustment for potential confounders, as appropriate.

Results

One hundred and twenty-six patients with ICI–inflammatory arthritis were included; 53 patients (42%) required a csDMARD/bDMARD. In the univariate logistic regression analysis, higher clinical disease activity index (CDAI), tenosynovitis, longer symptom duration before first rheumatology visit and longer ICI duration were significantly associated with a higher likelihood of requiring DMARDs; in addition, there was a trend towards those treated with prior chemotherapy being less likely to need DMARDs. After adjustment, tenosynovitis, longer symptom duration and higher CDAI remained associated with requiring DMARDs, while those with prior chemotherapy were significantly less likely to require DMARDs. Combination anti-CTLA-4 (cytotoxic T-lymphocyte-associated protein-4)/PD-1 (Programmed cell death protein-1) therapy and CS use at baseline were associated with a higher risk of persistent inflammatory arthritis.

Conclusion

Higher levels of disease activity, tenosynovitis and longer symptom duration prior to rheumatology referral were associated with requiring DMARDs for ICI–inflammatory arthritis, while those treated previously with chemotherapy were less likely to require additional immunosuppression. The presence of risk factors for severe disease at baseline may indicate a role for higher initial CS dose, earlier rheumatology referral, and adoption of immunosuppression beyond CSs to improve outcomes.

Rheumatology key messages.

• Many patients referred to rheumatology for ICI–inflammatory arthritis require DMARDs and have persistent arthritis.

• Higher disease activity, tenosynovitis, and delay in referral may increase the likelihood of requiring DMARD therapy.

• Persistent ICI–inflammatory arthritis was most common in those treated with combination ICI therapy, as previously shown.

Introduction

Immune checkpoint inhibitors (ICIs) are used across disparate types of cancer and improve outcomes for many patients [1]. ICIs can cause immune-related adverse events (irAEs), which occur due to excess immune activation and inflammation affecting non-cancer tissue [2]. Rheumatologists are referred patients on ICI therapy with new-onset inflammatory arthritis (ICI–inflammatory arthritis), PMR, myositis and other syndromes that resemble classic rheumatic diseases [3, 4]. ICI–inflammatory arthritis can cause significant impacts on patients’ health-related quality of life [5] and may lead to erosive disease [6]. The clinical spectrum of ICI–inflammatory arthritis is broad, and patients may differ by type of joint involvement, severity of symptoms and persistence of symptoms after ICI cessation [7].

A prior study identified increased duration of ICI therapy, presence of other irAEs and combination cytotoxic T-lymphocyte-associated protein-4 (CTLA-4)/programmed cell death protein-1 (PD-1) blockade as potential risk factors for the development of chronic ICI–inflammatory arthritis [8]. This study did not evaluate predictors for other markers of severe disease, such as requiring immunosuppression beyond CSs or developing structural damage. Understanding which patients may have an increased risk for severe outcomes serves several purposes in advancing our understanding of ICI–inflammatory arthritis. First, if duration or type of immunotherapy regimen, prior cancer treatments, or type of cancer predicted severe outcomes for ICI–inflammatory arthritis, consideration could be given to earlier use of immunomodulatory medications for prevention of inflammatory arthritis development in those patients. Second, clinical features that portend poor outcomes can be considered when deciding the intensity of first-line immunosuppression and frequency of monitoring for patients. Finally, in a clinical syndrome with much heterogeneity, defining subgroups has utility for future translational investigations. For example, in one study, CS-resistant ICI–inflammatory arthritis patients were those more likely treated with combination ICI therapy (Anti-PD-1/CTLA-4); these patients were found to have an enhanced Th-17 response [9]. The same differences in T cell activation could be examined for clinical subgroups based on other arthritis severity features.

Our aim was to evaluate whether features of antecedent cancer treatment, type and duration of immunotherapy, or clinical features of inflammatory arthritis at baseline affected the likelihood of requiring systemic CSs, requiring CS-sparing immunosuppression, or of having persistent inflammatory arthritis 6 months or more after ICI cessation.

Patients and methods

Study design

Participants were drawn from a prospective longitudinal study of patients with rheumatic irAEs (approved by the Johns Hopkins Institutional Review Board, #00263113) at Johns Hopkins; all participants gave informed consent. Patients were included if they had inflammatory arthritis diagnosed by a board-certified rheumatologist based on appropriate history and a physical examination showing synovitis, tenosynovitis, enthesitis and/or dactylitis and entered the longitudinal study prior to 1 April 2022. The patients were 18 years or older and had been treated with anti-PD-1/PD-L1 or anti-CTLA-4 therapy, either alone or in combination with other forms of therapy such as chemotherapy or relatlimab. Patients were excluded if they had prior systemic autoimmune disease.

Clinical data collection

Information on oncologic history, prior and current cancer treatments, and personal and family history of autoimmune disease were collected. Patients filled out surveys on global assessment of disease activity (100-point visual analog scale (VAS)), pain (100-point VAS) and stiffness severity (100-point VAS) at each visit. Physician-reported data included laboratory studies, imaging findings, 28 and 48 swollen and tender joint counts, global rating of disease severity (100-point VAS), use of CSs, conventional synthetic DMARDS (csDMARDs), and biologic DMARDs (bDMARDs), and cancer outcomes. Erosions were assessed by clinically obtained imaging studies, which included radiographs, USs and MRI studies.

Statistical analysis

Descriptive statistics were calculated for demographic and clinical features of the whole cohort and for the two groups by primary outcome. The primary binary outcome was requirement of either csDMARDs and/or bDMARDs specifically for ICI–inflammatory arthritis (determined by failure to improve on CSs in 6 weeks or failure to wean to <10 mg prednisone/day in 6 weeks) during follow up; the secondary outcomes were persistence of inflammatory arthritis at least 6 months after ICI cessation and requirement of systemic CSs at any point during follow up. T-tests or Wilcoxon rank sum tests, depending on the distribution of the data, and χ² tests were used to compare clinical characteristics by outcome categories. Logistic regression models were used to evaluate univariate associations between clinical features and the primary and secondary outcomes. Multivariable logistic regression models were run for the primary and secondary outcomes, including variables with P-values for associations of <0.10 in univariate models and with adjustment for potentially confounding factors, including age, sex, and CS use at baseline visit.

Results

Demographics and oncologic history

One hundred twenty-six patients with ICI–inflammatory arthritis were included in the study. The average age was 60 (s.d. 13) years; the majority were female (N = 70, 56%). The most common cancers were melanoma and lung cancer (Table 1). Patients were primarily treated with anti-PD-(L)1 agents (N = 89, 71%) or combination anti-CTLA-4/PD-1 therapy (N = 31, 26%). Most patients (58%) never smoked. Sixty-four patients (53.3%) were previously treated with chemotherapy, while 40 (32%) were previously treated with radiation.

Table 1.

Demographic features, cancer history, and selected arthritis variables by DMARD requirement

Variable	Total (n = 126)	No DMARD (n = 73)	csDMARD and/or bDMARD (n = 53)	P-value
Age, mean (s.d.)	60.0 (13.0)	60.0 (12.9)	60.1 (13.3)	0.98
Female gender, N (%)	70 (56%)	37 (51%)	33 (62%)	0.20
Race, N (%)				0.10
White	116 (92%)	65 (89%)	51 (96%)
Black	6 (5%)	6 (8%)	0 (0%)
Other	4 (3%)	2 (3%)	2 (4%)
Smoking status				0.81
Never	66 (57.9%)	37 (56.9%)	29 (59.2%)
Ever	48 (37.7%)	28 (43.1%)	20 (40.8%)
ICI class				0.14
Anti-PD-1/PD-L1	89 (70.6%)	52 (71.2%)	37 (69.8%)
Anti-CTLA-4	4 (3.2%)	4 (5.5%)	0 (0%)
Combination	31 (24.6%)	15 (20.6%)	16 (30.2%)
Other	2 (1.6%)	2 (2.7%)	0 (0%)
Cancer type				0.10
Melanoma	41 (32.5%)	18 (24.7%)	23 (43.4%)
Lung cancer	30 (23.8%)	20 (27.4%)	10 (18.9%)
GU cancer	7 (5.6%)	6 (8.2%)	1 (1.9%)
GI cancer	17 (13.5%)	13 (17.8%)	4 (7.6%)
Breast cancer	6 (4.8%)	2 (2.7%)	4 (7.6%)
SCC	8 (6.4%)	4 (5.5%)	4 (7.6%)
Other	17 (13.5%)	10 (13.7%)	7 (13.2%)
ICI duration at baseline (in months)	9.7 (9.4)	7.9 (8.4)	12.1 (10.3)	0.014
Prior chemotherapy, N (%)				0.08
No	56 (46.7%)	28 (40%)	28 (56%)
Yes	64 (53.3%)	42 (60%)	22 (44%)
Prior radiation, N (%)				0.65
No	86 (68%)	51 (69.9%)	35 (66%)
Yes	40 (32%)	22 (30.1%)	18 (34%)
Additional irAEs				0.12
0	76 (60.3%)	48 (65.8%)	28 (52.8%)
1	31 (24.6%)	18 (24.7%)	13 (24.5%)
2+	19 (15.1%)	7 (9.6%)	12 (22.6%)
Symptom duration at baseline (in months)	12.7 (10.0)	10.3 (7.5)	16.0 (12)	0.001
Tender joint count (28)	3.4 (4.0)	2.7 (4.0)	4.3 (3.9)	0.024
Swollen joint count (28)	7.0 (5.5)	6.1 (5.1)	8.2 (5.8)	0.034
Patient global disease activity VAS	39.7 (27.2)	37.7 (26.4)	42.4 (28.4)	0.35
Physician global disease activity VAS	25.5 (15.7)	22.4 (12.9)	29.5 (18.1)	0.02
CDAI	17.0 (10.3)	14.8 (8.8)	19.8 (11.4)	0.008
Pain VAS	48.3 (29.4)	48.9 (28.5)	47.4 (30.7)	0.78
Stiffness Severity VAS	52.2 (27.2)	51.1 (27.8)	53.5 (26.7)	0.65
Tenosynovitis Y/N	23 (18.9%)	7 (10%)	16 (30.8%)	0.004
Enthesitis Y/N	28 (22.6%)	19 (26.8%)	9 (17.0%)	0.20
Dactylitis	5 (4.0%)	2 (2.8%)	3 (5.8%)	0.40
Stenosing tenosynovitis	13 (10.7%)	5 (7.3%)	8 (15.1%)	0.16
Inflammatory back pain Y/N	3 (2.4%)	3 (4.2%)	0 (0%)	0.13
RF positive	12(10.4%)	8 (11.8%)	4 (8.5%)	0.58
Anti-CCP positive	7 (6.1%)	5 (7.8%)	2 (4.0%)	0.40
ANA positive	27 (24.8%)	15 (23.4%)	12 (26.7%)	0.70

Bold text highlights significant values. csDMARD: conventional synthetic DMARD; bDMARD: biologic DMARD; ICI: immune checkpoint inhibitor–induced; GU: genitourinary; GI: gastrointestinal; SCC: squamous cell carcinoma; irAE: immune-related adverse event; VAS: visual analogue scale; CDAI: clinical disease activity index; PD-L1: programmed cell death ligand-1; CTLA-4: cytotoxic T-lymphocyte-associated protein-4; PD-1: Programmed cell death protein-1.

Clinical features of inflammatory arthritis, treatment requirements, and outcomes

The average clinical disease activity index (CDAI) at baseline was 17 (s.d. 10.3). Tenosynovitis was present in 23 participants (18.9%), while enthesitis was present in 28 participants (22.6%). Stenosing tenosynovitis (e.g. trigger finger) was present in 13 participants (10.7%), while inflammatory back pain was rare, being present in only 3 participants (2.4%).

Of the 126 patients, 44 (34.9%) required a csDMARD, 25 (20%) were treated with a bDMARD, and 53 (42%) needed either a csDMARD or a bDMARD; requiring a csDMARD and/or a bDMARD was the primary outcome for the study. The most common csDMARD used was MTX (N = 26), followed by HCQ (N = 15) and LEF (N = 6). The most common class of biologic DMARD was TNF-inhibitors (N = 22), followed by IL-6R inhibitors (N = 3).

One hundred patients (80%) required systemic CSs at some point in their disease course. The average maximum CS treatment dose was 31.2 mg (s.d. 51.8) prednisone equivalent daily. Of the 126 patients, 114 had follow-up with rheumatology at least 6 months after ICI cessation, and 89 of those patients (78%) had persistent symptoms. Only 46 patients had imaging available after starting therapy for ICI-induced inflammatory arthritis, with 5 patients having erosions on follow-up imaging. Given the small sample with imaging, erosive disease was not further evaluated.

Predictors of primary outcome

Patients requiring a csDMARD and/or a bDMARD had a longer duration of ICI use, longer inflammatory arthritis symptom duration at first rheumatology visit, more tenosynovitis and higher disease activity (CDAI, tender joint count, swollen joint count, physician global assessment) than patients who did not require DMARDs (Table 1).

In univariate logistic regression analysis, there was a trend towards those treated with prior chemotherapy for cancer being less likely to need DMARDs. In contrast, in those patients with two or more irAEs, there was a trend towards being more likely to need DMARDs (Supplementary Table S1, available at Rheumatology online).

When adjusted for age, sex and CS use at baseline, it was found that tenosynovitis (odds ratio (OR): 3.80, 95% CI: 1.05, 13.76), higher CDAI (OR: 1.06, 95% CI: 1.01, 1.11) and longer symptom duration (OR: 1.06, 95% CI: 1.01, 1.12) remained significantly associated with requiring DMARDs, and prior chemotherapy was associated with 66% lower odds of DMARD use (Table 2). When biologic DMARD requirement alone was specifically evaluated as an outcome, the results did not differ from combining csDMARDs and bDMARDs (data not shown).

Table 2.

Unadjusted and adjusted OR for association with primary and secondary outcomes

Primary outcome: requiring a csDMARD or bDMARD
	Unadjusted OR (95% CI)	P-value	Adjusted OR (95% CI)	P-value
Age	1.00 (0.97, 1.03)	0.977	0.99 (0.96, 1.03)	0.931
Female sex	1.61 (0.78, 3.30)	0.198	1.31 (0.52, 3.40)	0.56
Prior chemotherapy	0.52 (0.25, 1.09)	0.085	0.34 (0.14, 0.85)	0.022
Already on CSs at baseline visit	1.99 (0.89, 4.44)	0.094	1.97 (0.73, 5.31)	0.182
Presence of tenosynovitis	4.00 (1.50, 10.63)	0.005	3.80 (1.05, 13.76)	0.042
CDAI	1.05 (1.01, 1.09)	0.01	1.06 (1.01, 1.11)	0.017
ICI duration	1.05 (1.01, 1.10)	0.02	1.03 (0.98, 1.09)	0.30
Symptom duration	1.06 (1.02, 1.11)	0.004	1.06 (1.01, 1.12)	0.027

Secondary outcome: persistent inflammatory arthritis >6 months after ICI cessation
	Unadjusted OR	P-value	Adjusted OR	P-value
Age	0.99 (0.96, 1.03)	0.593	0.98 (0.93, 1.02)	0.277
Female sex	1.27 (0.52, 3.08)	0.600	0.61 (0.18, 2.03)	0.419
Already on CSs at baseline visit	5.28 (1.16, 23.96)	0.031	5.51 (1.02, 29.90)	0.048
Combination ICI therapy	5.28 (1.16, 23.96)	0.031	5.43 (1.04, 28.43)	0.045
Presence of tenosynovitis	6.43 (0.81, 50.71)	0.078	7.09 (0.77, 65.16)	0.083
ICI duration	1.07 (0.99, 1.14)	0.065	1.08 (0.99, 1.18)	0.068
Symptom duration	1.13 (1.03, 1.24)	0.009	1.11 (0.99, 1.24)	0.067

Bold text highlights significant values. OR: odds ratio; cs DMARD: conventional synthetic DMARD; bDMARD: biologic DMARD; ICI: immune checkpoint inhibitor–induced; CDAI: clinical disease activity index.

Predictors of secondary outcomes

Persistence of ICI–inflammatory arthritis for 6 months or more after ICI cessation was next evaluated with univariate logistic regression. Longer symptom duration, being on CSs at the baseline visit, and combination ICI therapy were all significantly associated with higher odds of persistence; while not statistically significant, there were also trends that longer ICI treatment duration and the presence of tenosynovitis may be associated with higher odds of persistence (Table 2). In contrast to the findings for the primary outcome, those patients with prior use of chemotherapy did not have significantly lower odds of persistent inflammatory arthritis (OR = 0.64, 95% CI: 0.25, 1.64, Supplementary Table S2, available at Rheumatology online). CDAI at baseline was not associated with a higher likelihood of persistence (OR 1.04, 95% CI: 0.99, 1.10, Supplementary Table 2, available at Rheumatology online). Notably, RF or anti-CCP positivity were not associated with having persistent ICI–inflammatory arthritis (Supplementary Table 2, available at Rheumatology online).

In a multivariable model that also included age and sex, persistence remained significantly associated with being on CSs at baseline (OR 5.51, 95% CI: 1.03, 29.90) and receiving combination ICI therapy (OR 5.43, 95% CI: 1.04, 28.43), while there were trends towards tenosynovitis, longer symptom duration and longer ICI duration being associated with higher odds of persistence (Table 2).

Most patients (N = 108, 85.7%) required CSs at some point for treatment of ICI-induced inflammatory arthritis. Most variables evaluated were not significantly associated with a requirement for CSs in univariate analyses (Supplementary Table S3, available at Rheumatology online). Having RF was negatively associated with needing CSs (OR 0.17, 95% CI: 0.04, 0.68), and a higher CDAI at baseline was associated with a higher odds of CS use (OR 1.11, 95% CI: 1.02, 1.20).

Prior CS exposure

For the 62 patients reporting systemic CS use since their start of ICI therapy but prior to their baseline visit with rheumatology, we evaluated the relationship between maximum prior dose of CSs and primary and secondary outcomes with simple logistic regression. There were no differences between those treated with low-dose (prednisone or equivalent at ≤10 mg/daily), moderate-dose (10 mg < prednisone daily equivalent ≤ 60 mg), or high-dose (prednisone daily equivalent of >60 mg) CSs in either requiring a csDMARD/bDMARD or having persistent ICI–inflammatory arthritis (data not shown).

Discussion

In this study of 126 patients with ICI–inflammatory arthritis, higher CDAI, longer symptom duration and presence of tenosynovitis at the baseline rheumatology visit were associated with higher odds of requiring immunosuppression beyond CSs for ICI–inflammatory arthritis; these associations persisted after adjusting for demographic features, prior chemotherapy, and baseline CS use. Conversely, those treated previously with chemotherapy were less likely to require additional immunosuppression. There were similar findings with the secondary outcomes; additionally, receiving combination anti-CTLA-4/anti-PD-1 therapy as opposed to another ICI regimen was associated with persistence of inflammatory arthritis longer than 6 months after ICI cessation.

Though ICI–inflammatory arthritis is known to be a heterogeneous disease, this study was the first to evaluate features at the baseline rheumatology visit for predicting severe outcomes. It follows our group’s previous work on the persistence of ICI–inflammatory arthritis, in which ICI duration, combination ICI therapy, and having other irAEs were associated with a higher likelihood of persistence [8]. Interestingly, combination therapy was not associated with higher odds of receiving csDMARD/bDMARD therapy for ICI–inflammatory arthritis, and, conversely, some baseline features associated with csDMARD/bDMARD requirement (such as CDAI) were not associated with persistence. These differing associations between the outcomes may reflect differences in underlying biology between those with severe disease who require DMARDs and those who go on to develop persistence of inflammatory arthritis.

In addition, this study reinforced findings about ICI–inflammatory arthritis phenotype and severity. Enthesitis and tenosynovitis were common (∼20% for each) as reported in prior studies using US imaging [6]. The tendon predominance, which is similar to forms of SpA, may be related to the Th17 immune response seen in some patients with ICI–inflammatory arthritis [9]. A systematic literature review showed that ICI–inflammatory arthritis is primarily seronegative and that most patients referred to rheumatology require systemic CSs for treatment, findings which were also redemonstrated [7].

Prior studies have examined risk factors for developing ICI–inflammatory arthritis, regardless of severity. In one study, having melanoma or genitourinary cancer vs lung cancer and being treated with combination anti-PD-1/anti-CTLA-4 therapy were associated with higher likelihood of developing ICI–inflammatory arthritis [10]. Studies that have examined irAEs more broadly have found a variety of other risk factors, including antibiotic use [11]. Similar to our findings on ICI duration, a study of multisystem irAEs showed that longer ICI duration along with favourable performance status were independent risk factors for development of multisystem irAEs [12]. In a study of irAEs of all types, poor performance status, elevated neutrophil/lymphocyte ratio, and lung cancer were associated with higher rates of grade IV or V irAEs [13]. Overall, risk factors seem to differ by type of irAE and by type of agent or cancer studied, with some studies showing opposite results, as with performance status, and with no prevailing shared risk factors yet discovered [14].

Limitations of the study include being a single-centre cohort with primarily older patients and limited racial diversity. Multiple tumour types and ICI regimens were represented, which suggests that the results are generalizable among the adult oncology population. Another limitation was the lack of serial imaging in most patients to evaluate for structural changes like erosions, another important marker of inflammatory arthritis severity.

These findings can inform management decisions and clinical trial design for ICI–inflammatory arthritis. Those patients with moderate to severe disease activity may be better served by early csDMARD/bDMARD therapy to limit CS dose if they will eventually require these agents. Similarly, patients who are likely to have persistent inflammatory arthritis (for example, those on combination ICI therapy) may be prioritized for referral to rheumatologic care and may also benefit from earlier introduction of CS-sparing agents. Future studies should validate these findings in multicentre, diverse cohorts so that they may inform models of risk stratification and treatment decisions for ICI–inflammatory arthritis.

Supplementary material

Supplementary material is available at Rheumatology online.

Data availability

Data will be made available upon reasonable request to the authors.

Funding

Funding for this study was provided by the National Institute of Arthritis and Musculoskeletal and Skin Diseases (NIAMS) (K23 AR075872, K24 AR080217, P30 AR070254, R01 AR073208) and a Rheumatology Research Foundation K Supplement.

Disclosure statement: L.C.C. has received research funding from Bristol-Myers Squibb. O.K. and M.J. have no conflicts of interest to report. C.O.B. has received research funding from Bristol-Myers Squibb and has been a consultant for AbbVie, Eli Lilly, Janssen, Pfizer, and Sanofi. A.A.S. has received research funding from Arena Pharmaceuticals, Medpace LLC, Eicos Sciences, and the Kadmon Corporation.