Predictors of Immunosuppressive Treatment in Cutaneous Arteritis: Analysis of Clinical Characteristics in 42 Patients

ABSTRACT

Cutaneous arteritis is a necrotizing vasculitis of medium‐sized vessels in the skin for which no established criteria exist to guide the initiation of systemic treatment. This study aimed to identify the clinical and laboratory markers associated with the initiation of immunosuppressive treatments, including systemic corticosteroids. This retrospective chart review included 42 patients with cutaneous arteritis treated at the Department of Dermatology, Yokohama City University Hospital, from 2010 to 2023. Clinical manifestations and laboratory and imaging data were collected. Univariable and multivariable analyses were performed to identify the risk factors associated with the initiation of immunosuppressive treatment. Twenty‐five patients with cutaneous arteritis (five men, 20 women) were included in this study. Of these, 14 patients received systemic corticosteroids and 11 were treated with additional immunosuppressive drugs. The Youden index analysis indicated that the thresholds of systemic corticosteroid induction were patients older than 43 years and platelet counts more than 340 000/μL. Univariable Firth logistic regression analysis showed numbness in the lower limbs, older age at onset, elevated platelet counts, and D‐dimer levels (≥ 1 μg/mL) as risk factors for systemic corticosteroids. Our analysis identified several risk factors, such as numbness in the lower limbs, that may help predict the need for immunosuppressive treatment. Further large‐scale prospective studies are required to confirm these findings.

1. Introduction

Cutaneous arteritis (CA) is a rare disease characterized by skin‐limited necrotizing vasculitis of medium‐sized vessels. Although CA was formerly known as cutaneous polyarteritis nodosa (PAN), it was recently defined as a single‐organ vasculitis at the revised Chapel Hill Consensus Conference (CHCC) in 2012 [1]. Generally, CA does not cause systemic symptoms, such as fever or other organ damage, as seen in PAN; however, limited systemic symptoms can sometimes be observed at the affected site, such as arthralgia and neuropathy. A recent study showed a 5‐year survival rate of 94.0% in PAN; however, there are no detailed reports on the survival rates in patients with CA [2]. In contrast to PAN, some cases of CA do not require aggressive immunosuppressive therapy and only improve with the administration of antithrombotic drugs (anticoagulants and antiplatelet drugs) [3]. According to the 2023 Japanese Guidelines for the Treatment of Cutaneous Vasculitis and Vasculopathy, nonsteroidal anti‐inflammatory drugs (NSAIDs) and antithrombotic drugs are recommended for mild cases. Systemic corticosteroids should be administered in cases of refractory or severe cutaneous symptoms such as skin ulcers and gangrene, and additional immunosuppressive drugs should be considered in cases of poor response [4]. However, no randomized controlled trials or observational studies have directly compared the treatments for CA, and no standardized treatment protocol has been established.

In this study, we retrospectively analyzed patients with CA treated at our hospital to identify risk factors associated with the induction of immunosuppressive treatments. Furthermore, the Youden index was used to calculate the cutoff values for the factors predicting the initiation of systemic corticosteroid or immunosuppressive therapy.

2. Methods

2.1. Subjects

Patients with CA treated at Yokohama City University Hospital between January 2010 and September 2023 were included in this study. The patients with CA fulfilled the 2012 updated Chapel Hill Consensus Conference [3] and Nakamura criteria [5]. Biopsy specimens were obtained at sufficient depth to evaluate medium‐ and small‐sized arteries, and serial sections were performed. When the initial specimen was considered insufficient, repeat biopsies were obtained when needed. This study was approved by the Institutional Review Board of Yokohama City University (approval number: F230100019). This study was conducted using an opt‐out approach. Information regarding the study was disclosed on the institutional website, and participants were given the opportunity to decline participation.

2.2. Clinical and Biological Data

The collected data included sex, age at disease onset, duration of symptoms before diagnosis, and clinical findings, such as erythema, purpura, livedo racemosa, pigmentation, subcutaneous nodules, skin ulcers, arthralgia, numbness in the lower limbs, and mononeuritis multiplex. Treatment history, including antithrombotic drugs, systemic corticosteroids, and immunosuppressive drugs, and duration of treatment, were also recorded. The presence of lower limb numbness was assessed using medical records by integrating patient‐reported symptoms with the findings of neurological examinations performed by neurologists. Patients presenting with lower limb numbness were referred to neurologists for further evaluation. Laboratory findings included complete blood count, erythrocyte sedimentation rate (ESR), C‐reactive protein (CRP) level, creatinine (Cr) level, D‐dimer level, neutrophil count, platelet count, and neutrophil‐to‐lymphocyte ratio. Blood tests were performed before treatment initiation. Exacerbation was defined as the appearance of new rashes, subjective symptoms such as pain or numbness, or the worsening of these symptoms. Systemic corticosteroids were initiated when any of the following conditions were observed despite antithrombotic therapy: (i) new cutaneous manifestations (e.g., erythema, subcutaneous nodules, or skin ulcers); (ii) persistent pain or neurological symptoms without improvement; or (iii) clear exacerbation of existing skin lesions. Relapse was defined as the appearance of new lesions or the worsening of existing symptoms. The addition of immunosuppressive agents was considered in such cases.

2.3. Statistical Analysis

All continuous variables are expressed as mean ± standard deviation. The two groups were compared using Fisher's exact test and the Mann–Whitney U test. Some factors were log‐transformed because they were skewed toward the high side (Figure S1). For comparisons between more than two groups, a one‐way analysis of variance with Tukey's multiple comparison test was used. The cutoff values were calculated using the Youden Index. To assess the robustness of the univariable analysis and address confounding factors, bivariate logistic regression models were applied to all possible combinations of the following risk factors: sex, age of onset, duration of symptoms before diagnosis, erythema, purpura, livedo racemosa, pigmentation, subcutaneous nodules, skin ulcers, arthralgia, numbness in the lower limbs, mononeuritis multiplex, ESR, CRP, Cr, D‐dimer, neutrophil count, platelet count, and neutrophil‐to‐lymphocyte ratio. Several variables in this study could have acted as potential confounders; however, the limited number of outcome events precluded the use of a reliable multivariable model. Therefore, only univariable and bivariable analyses were performed. Additionally, Firth's correction was applied to address the separation issues, as shown in Table S6 and Figure S2. Univariable Firth logistic regression analyses were first conducted with systemic corticosteroid induction and immunosuppressive drug induction as response variables. Bivariable analyses were then performed by considering combinations of relevant risk factors (see the Discussion section for details). All statistical analyses were conducted using GraphPad Prism software (version 8.2; GraphPad Software, San Diego, CA, USA), JMP Pro software (version 16.1.0; SAS Institute Inc., Cary, NC, USA), and R software (version 4.3.1; R Foundation for Statistical Computing, Vienna, Austria). Statistical significance was set at p < 0.05.

3. Results

3.1. Study Selection and Patient Characteristics

Initially, 42 patients who were clinically treated for CA at Yokohama City University Hospital were enrolled (Figure 1). Among them, 11 patients were excluded because pathological examination failed to reveal necrotizing vasculitis in the small to medium‐sized arteries. An additional two patients were lost to follow‐up for personal reasons, and four patients were excluded because of worsening underlying disease, which made the evaluation of CA progression difficult. Consequently, 25 patients with at least 6 months of clinical follow‐up after treatment initiation were included in this study. To assess potential selection bias, we compared the baseline characteristics of included and excluded patients, and no statistically significant differences were observed between the two groups (Table S1).

FIGURE 1.

Flowchart of study design. A flowchart of the study design is shown. Patients with cutaneous arteritis (CA) were divided into groups based on treatment with or without systemic corticosteroids and immunosuppressive drugs.

The clinical characteristics are summarized in Table 1. The median age was 50 years (range: 29–60 years). The cohort comprised five men (20.0%) and 20 women (80.0%). The mean age of CA onset was 44.0 (range, 26–54 years), and the mean duration from symptom onset to diagnosis was 22.3 ± 37.2 months. The most common clinical manifestation was erythema (92%, n = 23), followed by livedo racemosa and pigmentation (48%, n = 12). Other clinical findings included subcutaneous nodules (36%, n = 9), arthralgia (36%, n = 9), numbness in the lower legs (36%, n = 9), and skin ulcers (12%, n = 3). All nine patients with lower limb numbness were comprehensively evaluated by neurologists and underwent nerve conduction studies and electromyography. Among these, four patients were diagnosed with mononeuritis multiplex. For the remaining five patients, the neurologists determined that their numbness was most likely attributable to cutaneous arteritis, with alternative etiologies, such as diabetic neuropathy and spinal disorders, deemed clinically unlikely.

TABLE 1.

Clinical characteristics and laboratory findings of 25 patients with cutaneous arteritis.

Variable	Total (n = 25)
Gender, Male, n (%)	5 (20.0)
Age (years), median (IQR)	50 (29–60)
Age of CA onset (years), median (IQR)	44 (26–54)
Disease duration (months), mean ± SD	22.3 ± 37.2
Follow‐up duration (months), mean ± SD	64.9 ± 46.1
Cutaneous manifestation
Erythema, n (%)	23 (92.0)
Purpura, n (%)	2 (8.0)
Livedo racemosa, n (%)	12 (48.0)
Pigmentation, n (%)	12 (48.0)
Nodule, n (%)	9 (36.0)
Skin ulcer, n (%)	3 (12.0)
Laboratory findings
Plt (104/uL), mean ± SD	28.2 ± 8.6
Neu (102/uL), mean ± SD	18.1 ± 42.0
NLR (%), mean ± SD	3.8 ± 3.6
CRP (mg/dl), mean ± SD	1.1 ± 0.8
Cr (mg/dl), mean ± SD	0.6 ± 0.1
ESR (mm/h), mean ± SD	18.3 ± 21.9
D‐dimer > 1 μg/mL, n (%)	5 (20.0)
Complication, n (%)
Arthralgia, n (%)	9 (36.0)
Numbness in the lower leg, n (%)	9 (36.0)
Mononeuritis multiplex	4 (16.0)
Treatment
Antithrombotic drugs	25 (100)
Systemic corticosteroids, n (%)	14 (56.0)
Immunosuppressive drugs, n (%)	10 (40.0)
Azathioprine	9 (36.0)
Mizoribine	3 (12.0)
Methotrexate	3 (12.0)

Abbreviations: CRP, C‐reactive protein; ESR, erythrocyte sedimentation rate; IQR, interquartile range; Neu, neutrophils; NLR, neutrophil‐to‐lymphocyte ratio; Plt, platelets; SD, standard deviation.

In laboratory findings, mean CRP (1.17 ± 0.82 mg/dL) and ESR (18.33 ± 21.95) were higher than normal. Five patients (20%) had elevated D‐dimer levels more than 1 μg/mL. All patients initially received antithrombotic drugs during their first visit. Antithrombotic drugs were continued in all patients after the initiation of systemic corticosteroids or immunosuppressive agents. The types of antithrombotic agents used in the steroid and non‐steroid groups are summarized in Table S7. Subsequently, 14 patients required systemic corticosteroids (prednisolone 0.5 mg/kg/day) because of inadequate efficacy. Eleven of these patients later received immunosuppressive drugs because of relapse or lack of improvement. The median time from diagnosis to initiation of systemic corticosteroid treatment was 70.0 days (interquartile range (IQR): 51–86). Among patients who subsequently required immunosuppressive therapy, the median time from initiation of corticosteroids to addition of immunosuppressants was 116.7 days (IQR 73–123).

3.2. Risk Factors for Systemic Corticosteroid Induction

To identify the risk factors requiring systemic corticosteroids, we compared patients treated with antithrombotic drugs only (n = 11, non‐steroid group) with those who received systemic corticosteroids (n = 14, steroid group) (Table 2). The mean age at onset was significantly higher in the steroid‐treated group than in the non‐steroid‐treated group (p = 0.038). Furthermore, the proportion of patients with lower leg numbness and D‐dimer levels > 1 μg/mL in the steroid group was significantly higher than that in the non‐steroid group (p < 0.001 and p = 0.029, respectively). Platelet count was also higher in the steroid‐treated group (p = 0.006). We further investigated the utility of the platelet count and age at onset as biomarkers for systemic corticosteroid induction. Using the Youden index, we established the optimal cutoff values: age at onset > 43 years (sensitivity, 78.6%; specificity, 72.7%) and platelet count > 340 000/μL (sensitivity, 64.3%; specificity, 90.9%). The corresponding area under the curve (AUC) values were 0.730 (95% CI, 0.572–0.890) and 0.857 (95% CI, 0.620–0.957), respectively. Univariable Firth logistic regression analysis showed that age at onset (OR 1.04, 95% CI 1.00–1.11, p = 0.038), lower limb numbness (OR 39.73, 95% CI 3.82–5480.44, p < 0.001), platelet count (OR 1.16, 95% CI 1.04–1.34, p = 0.007), and D‐dimer elevation (OR 13.32, 95% CI 1.24–1833.62, p = 0.030) were significantly associated with the initiation of systemic corticosteroids (Table S2). To explore potential confounding, bivariable logistic regression analyses were performed. Bivariable analyses demonstrated that these associations remained robust after adjustment for potential confounders (Table S3).

TABLE 2.

Comparison of clinical characteristics between the non‐steroid and steroid groups.

Variable	Non‐steroid (n = 11)	Steroid (n = 14)	p
Gender, Male, n (%)	1 (9.0)	4 (28.5)	0.266
Age (years), median (IQR)	40 (26–49.5)	55.5 (47.5–61.5)	0.063
Age of CA onset (years), median (IQR)	29 (21–41.5)	49.5 (43.2–58.7)	0.038
Disease duration (months), mean ± SD	16.3 ± 18.4	27.0 ± 48.5	0.301
Follow‐up duration (months), mean ± SD	85.8 ± 17.8	48.5 ± 7.2	0.046
Cutaneous manifestation
Erythema, n (%)	10 (90.9)	13 (92.8)	0.837
Purpura, n (%)	1 (9.0)	1 (7.1)	0.837
Livedo racemosa, n (%)	7 (63.6)	5 (35.7)	0.179
Pigmentation, n (%)	4 (36.3)	8 (57.1)	0.320
Nodule, n (%)	3 (27.2)	6 (42.8)	0.446
Skin ulcer, n (%)	0 (0)	3 (21.4)	0.138
Laboratory findings
Plt (104/uL), mean ± SD	23.8 ± 1.7	31.9 ± 2.3	0.006
Neu (102/uL), mean ± SD	35.5 ± 34.3	44.0 ± 59.9	0.520
NLR (%), mean ± SD	3.0 ± 0.6	4.1 ± 1.3	0.785
CRP (mg/dl), mean ± SD	0.3 ± 0.1	1.1 ± 0.4	0.430
Cr (mg/dl), mean ± SD	0.7 ± 0.03	0.7 ± 0.06	0.764
ESR (mm/h), mean ± SD	15.0 ± 3.6	25.9 ± 4.6	0.160
D‐dimer > 1 μg/mL, n (%)	0 (0)	5 (35.7)	0.029
Complication, n (%)
Arthralgia, n (%)	2 (18.1)	7 (50.0)	0.112
Numbness in the lower leg, n (%)	0 (0)	9 (64.2)	< 0.001
Mononeuritis multiplex	0 (0)	4 (28.5)	0.065
Treatment
Antithrombotic drugs	11 (100)	14 (100)	—
Systemic corticosteroids, n (%)	—	14 (100)	—
Immunosuppressive drugs, n	—	10	—
Azathioprine, n	—	9	—
Mizoribine, n	—	3	—
Methotrexate, n	—	3	—

Abbreviations: CRP, C‐reactive protein; ESR, erythrocyte sedimentation rate; IQR, interquartile range; NA, not applicable; Neu, neutrophils; NLR, neutrophil‐to‐lymphocyte ratio; Plt, platelets; SD, standard deviation.

These results suggest that factors related to inflammatory responses, along with lower limb numbness, are independent risk factors for the potential requirement for systemic corticosteroids.

We further analyzed the associations between cutaneous manifestations and systemic corticosteroid induction. Logistic regression analysis revealed that skin ulcers were notably associated with systemic corticosteroid induction in several two‐variable models. Other cutaneous manifestations did not show significant associations. Interestingly, skin ulcers were consistently associated with lower limb numbness in an exploratory two‐variable logistic regression analysis using lower limb numbness as the outcome.

3.3. The Risk of Additional Immunosuppressant Induction

We further investigated the risk factors associated with an additional requirement of immunosuppressants. Additional immunosuppressive drugs are generally required in patients with CA who experience symptom relapse, insufficient response to systemic corticosteroids, or difficulty in tapering corticosteroid doses. The patients were divided into two groups: 15 patients treated with antithrombotic drugs with or without systemic corticosteroids (non‐immunosuppressive therapy group), and 10 patients treated with systemic corticosteroids in combination with immunosuppressive drugs (immunosuppressive therapy group). Compared to the non‐immunosuppressive therapy group, the immunosuppressive therapy group had a significantly higher proportion of patients with numbness in the lower leg (p = 0.004), older age of CA onset (p < 0.001), and significantly higher platelet counts (p = 0.004) and ESR values (p = 0.046) (Table 3). Univariable Firth logistic regression analysis demonstrated that age at onset (OR 1.12, 95% CI 1.05–1.28, p < 0.001), lower limb numbness (OR 11.57, 95% CI 2.05–90.32, p = 0.005), platelet count (OR 1.16, 95% CI 1.04–1.34, p = 0.006), and log (ESR) (OR 2.59, 95% CI 1.02–8.92, p = 0.046) were significantly associated with the initiation of immunosuppressive therapy (Table S4). Bivariable analyses confirmed that these associations remained robust after adjustment for potential confounders (Table S5).

TABLE 3.

Comparison of clinical characteristics between non‐immunosuppressant and immunosuppressant groups.

	Non‐immunosuppressant (n = 15)	Immunosuppressant (n = 10)	p
Gender, Male, n (%)	2 (13.3)	3 (30.0)	0.323
Age (years), median (IQR)	34.0 (23.5–47.5)	60.0 (54.7–70.2)	< 0.001
Age of CA onset (years), median (IQR)	29.0 (20.5–40.0)	54.5 (48.7–66.7)	< 0.001
Disease duration (months), mean ± SD	13.4 ± 4.2	35.8 ± 17.6	0.886
Follow‐up duration (months), mean ± SD	70.8 ± 14.6	56.0 ± 8.7	0.450
Cutaneous manifestation
Erythema, n (%)	13 (86.6)	10 (100)	0.342
Purpura, n (%)	2 (13.3)	0 (0)	0.342
Livedo racemosa, n (%)	8 (53.3)	4 (40.0)	0.531
Pigmentation, n (%)	7 (46.6)	5 (50.0)	0.873
Nodule, n (%)	4 (26.6)	5 (50.0)	0.248
Skin ulcer, n (%)	1 (6.6)	2 (20.0)	0.341
Laboratory findings
Plt (104/uL), mean ± SD	23.9 ± 1.5	33.6 ± 3.0	0.006
Neu (102/uL), mean ± SD	34.2 ± 26.8	49.4 ± 77.2	0.192
NLR (%), mean ± SD	2.8 ± 0.4	4.9 ± 1.8	0.246
CRP (mg/dl), mean ± SD	0.4 ± 0.2	1.3 ± 0.5	0.124
Cr (mg/dl), mean ± SD	0.6 ± 0.02	0.7 ± 0.07	0.419
ESR (mm/h), mean ± SD	15.7 ± 3.3	29.2 ± 5.5	0.046
D‐dimer > 1 μg/ml, n (%)	2 (13.3)	3 (30.0)	0.323
Complication, n (%)
Arthralgia, n (%)	5 (33.3)	4 (40.0)	0.730
Numbness in the lower leg, n (%)	2 (13.3)	7 (70.0)	0.004
Mononeuritis multiplex	1 (6.67)	3 (30.0)	0.141
Treatment
Antithrombotic drugs	15 (100)	10 (100)	—
Systemic corticosteroids, n (%)	4 (26.6)	10 (100)	—
Immunosuppressant, n (%)	—	10 (100)	—
Azathioprine, n (%)	—	9 (90.0)	—
Mizoribine, n (%)	—	3 (30.0)	—
Methotrexate, n (%)	—	3 (30.0)	—

Abbreviations: CRP, C‐reactive protein; ESR, erythrocyte sedimentation rate; IQR, interquartile range; NA, not applicable; Neu, neutrophils; NLR, neutrophil‐to‐lymphocyte ratio; Plt, platelets; SD, standard deviation.

4. Discussion

This study identified the potential risk factors for treatment escalation in patients with CA. Our analysis revealed that older age at CA onset, elevated platelet count, presence of numbness in the lower leg, and elevated D‐dimer levels were associated with the induction of systemic corticosteroids. Furthermore, additional induction of immunosuppressive drugs was found to be increased in patients with older age of CA onset, elevated platelet counts, and lower leg numbness. Given the current lack of clear criteria for immunosuppressive therapy for CA, our findings may provide an important basis for guiding critical treatment decisions. Our study also provided potential cutoff values for age and platelet count, calculated using the Youden index. These values could help predict when treatment should be initiated in clinical practice. However, given the small sample size, these cut‐off values should be considered exploratory and require external validation before clinical application.

Our results are consistent with the hypothesized role of thrombo‐inflammation in CA pathogenesis, as suggested in previous reports and guidelines [4, 6, 7, 8]. The Japanese Guidelines for the Treatment of Cutaneous Vasculitis and Vasculopathy (2023) recommend targeting circulatory dysfunction in mild CA cases [4], a concept supported by the efficacy of NSAIDs and warfarin [7]. Moreover, Kawakami et al. reported the detection of anti‐phosphatidylserine prothrombin complex (anti‐PS/PT) antibodies and/or lupus anticoagulant in 16 patients with CA, suggesting an underlying hypercoagulable state in this disease and its presence as a potential marker in patients with CA [8]. These reports led us to speculate that local circulatory dysfunction may trigger the development of CA. Our study extends these observations by demonstrating that patients resistant to initial antithrombotic drug monotherapy were characterized by older age, higher incidence of numbness, and significantly elevated platelet counts and D‐dimer levels. In severe CA, persistent vascular injury is likely to drive platelet activation and fibrin clot formation, leading to elevated D‐dimer levels. In addition, chronic exposure of platelets to inflammation exacerbates the inflammation, and there is no association with damaged blood vessels. This creates a vicious cycle of thrombo‐inflammation [9]. We hypothesized that this cycle renders antithrombotic monotherapy insufficient in high‐risk patients, necessitating systemic corticosteroids to suppress the overwhelming inflammatory response. Meanwhile, elevated platelet and D‐dimer levels may also contribute to the pathogenesis of CA, and future studies should verify the detailed mechanisms.

In previous studies, elevated CRP levels before treatment, presence of skin ulcers, and mononeuritis multiplex were identified as predictors of the induction of systemic corticosteroids [10, 11, 12, 13]. In our study, although CRP levels were higher in the steroid‐treated group than in the non‐steroid‐treated group, the difference was not statistically significant. However, the incidence of lower limb numbness was significantly higher in the steroid‐treated group than in the non‐steroid‐treated group. A similar trend was observed in the immunosuppressant and non‐immunosuppressant groups. Notably, the findings of the logistic regression analysis were consistent, suggesting that the presence of lower limb numbness is an important clinical symptom for predicting the induction of systemic corticosteroids and immunosuppressants. In vasculitis, the etiology of numbness is probably an ischemic infarction of the peripheral nerves by occlusion of the peripheral neurotrophic vessels. More serious vascular damage may occur in severe cases requiring systemic corticosteroids, resulting in ischemic infarction of the peripheral nerves. Consistent with this concept, logistic regression analysis showed that skin ulcers were associated with lower limb numbness. Although numbness does not necessarily indicate confirmed neuropathy in all patients, it may be a clinically significant symptom of peripheral neuritis. Therefore, early diagnosis of CA is important to prevent irreversible motor sensory impairment. However, numbness can result from various causes and is often not primarily due to nerve involvement, which makes differential diagnosis challenging. Isahaya et al. analyzed peripheral nerve involvement in 164 patients with CA; among them, 17 patients (10.4%) had neurological symptoms, 7 of whom (41.2%) were diagnosed with polycystic mononeuritis [14]. Interestingly, there was no clinically significant difference between the patients with mononeuritis multiplex and those with only neurological symptoms. This suggests that the presence of numbness is clinically important, even without a formal diagnosis of mononeuritis multiplex. As this was a retrospective study based on medical records, the possibility of residual confounding factors, including unrecognized causes of neuropathy, cannot be completely excluded.

In addition, our study identified older age at CA onset as an important predictor of immunosuppressive drug induction. In mild cases of CA, the absence of symptoms and delay in consulting a clinic or hospital may affect the outcomes. A questionnaire survey of cutaneous arteritis findings among dermatologists familiar with vasculitis certified by the Vasculitis and Vascular Disease Practice Guidelines Committee of the Japanese Dermatological Association showed that the mean age in the steroid‐treated group was significantly higher than that in the non‐steroid‐treated group [15]. This report supports our findings, while there was no association between age at disease onset and immunosuppressive treatment in previous reports [10, 11, 12, 13]. In our analysis, even in the logistic model adjusted for other clinical symptoms such as numbness in the lower limbs, older age at CA onset remained a significant risk factor. This suggests that although older age is an independent risk factor, its effect may be amplified by the coexistence of other complications, such as numbness, collectively contributing to a more severe and treatment‐resistant disease course.

Our study has some limitations. First, the number of enrolled patients with CA was small. This limitation makes it difficult to perform multivariate analysis and detect statistically significant differences or associations. To confirm the consistency of the univariable analysis, we used a brute‐force approach with a two‐variable model logistic regression analysis for the 19 candidate risk factors. Some factors were log‐transformed because they were skewed toward the high side (Figure S1). In addition, in some cases, the correlation between prognostic factors was strong, while in others, the estimates of the odds ratio in the 2 × 2 contingency table were unstable owing to insufficient data. Therefore, we applied the Firth correction to ensure more reliable estimates (Table S6). Although Firth's correction was applied to address the instability of estimates in a small sample, the reproducibility of our findings should be confirmed in larger cohorts. Because Figure S2 suggests a strong correlation between sex and platelet count, the results of logistic regression models including both variables as covariates may have been affected by multicollinearity; therefore, these findings were not included in the main interpretation. Second, this was a single‐center cohort study. Since university hospitals receive more patients with severe CA than clinics and general hospitals, there was a selection bias in the recruitment of the study population. Third, the variability in the definition of immunosuppressive treatment induction (treatment resistance) among different studies complicates the comparison and generalization of the present findings. Furthermore, differences in follow‐up duration between groups may have influenced the detection of relapse and subsequent initiation of immunosuppressive therapy.

In conclusion, our logistic regression analysis showed that risk factors such as the presence of numbness in the lower limbs, increased platelet count, and old age at CA onset were related to the induction of immunosuppressive treatment. These findings may help predict the need for immunosuppressive therapy. Future large‐scale, prospective multicenter studies are required to validate these clinical risk factors.

Ethics Statement

The study protocol was approved by the Institutional Review Board of Yokohama City University (approval number: F230100019).

Consent

This study was conducted using an opt‐out approach in accordance with the Ethical Guidelines for Medical and Health Research Involving Human Subjects in Japan. Information about the study was disclosed to all potential participants, and opportunities to decline participation were provided.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Table S1: baseline comparisons between included and excluded patients.

Tables S2–S5: detailed univariable and bivariable Firth logistic regression analyses.

Table S6: contingency table data supporting the use of Firth's correction.

Table S7: antithrombotic treatment details.

Figures S1 and S2: variable distributions before log transformation and the association between platelet count and gender.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.