Clinical characteristics and prognosis of castleman disease patients in a Chinese hospital: paraneoplastic pemphigus is an independent risk factor

Background: Castleman disease (CD) is a rare lymphoproliferative disorder that has had limited clinical research. This study aims to detect the clinical manifestations, pathological features, and prognostic factors of this disease. Methods: This study retrospectively analyzed the information of 54 patients with CD hospitalized in a single centre. A Cox regression model was employed to perform univariate analysis and multivariate analysis in order to identify independent prognostic factors for survival. Results: Based on clinical classication, 30 patients (55.6%) had unicentric CD (UCD) and 24 patients (44.4%) had multicentric CD (MCD). Moreover, pathological classication identied 32 cases (59.3%) with hyaline vascular variant (HV), 3 (5.6%) with mixed cellular variant (Mix), and 19 (35.2%) with plasmacytic variant (PC). The MCD patients commonly exhibited clinical signs and symptoms, including fever, splenomegaly, and pleural effusion and/or ascites. Several clinical complications, such as liver injury, anemia, and polyradiculoneuropathy, organomegaly, endocrinopathy, monoclonal plasma cell disorder, skin changes syndrome (POEMS) were more common in MCD patients. Univariate analysis showed that presence of paraneoplastic pemphigus (PNP) and elevated C-reactive protein (CRP) were unfavorable factors relating to CD patient survival. Multivariate analysis identied the presence of PNP as an independent prognostic factor in patients with CD. Conclusions: This study provided a panoramic elaboration of CD cases and showed the presence of PNP was an independent unfavorable factor.


Introduction
Castleman disease (CD) is a rare lymphoproliferative disorder that was rst described by Dr. Benjamin Castleman 60 years ago [1]. In recent decades, a number of case reports and reviews have presented the clinical manifestations [2,3], pathological features [4,5], clinical treatment [6], and have attempted to explain the pathogenesis [7] of this complicated disease. However, due to the low morbidity, the study on CD has progressed slowly.
Clinically, CD is characterized as unicentric (UCD) and multicentric (MCD) based on the centricity. UCD is typically localized without systemic involvement, thus surgery is the main treatment. On the contrary, MCD is a systemic disorder that comprises two subgroups: human herpesvirus 8 (HHV8)-related MCD [9,10], and idiopathic multicentric Castleman disease (iMCD) [11,12]. Systemic therapies are primarily applied to MCD. Based on pathology, CD can be classi ed into hyaline vascular variant (HV), plasmacytic variant (PC) and mixed cellular variant (Mix). A systematic study on 416 CD patients established a novel classi cation system that provided a valuable model for the prediction of midterm outcome [13].
Although a large number of studies have focused on the clinical and pathological features during occurrence and progression of CD, few studies have examined the risk factors that in uence CD prognosis. In this study, we reviewed a cohort of 54 patients with CD from a single center in China, in order describe the outcome of this disease with complex clinical manifestations, and de ne the prognostic factors.

Patient characteristics
We retrospectively collected the clinical and pathological data for 54 Chinese patients diagnosed with CD from 2008 to 2018 in The First A liated Hospital of Nanjing Medical University. The pathological data of each patient was based on the tissue specimens obtained from 23 needle biopsies and 31 surgical excisions, and were reviewed by at least two experienced pathologists Fig 1 . The pathological classi cation of CD was established according to Cronin and Kellers criteria [4,5]. The clinical classi cation of all enrolled cases was based on physical and imageological examination. Other de nitions in this study included: (1) anemia, de ned as hemoglobin < 110 g/l for females and < 120 g/l for males; (2) hypoalbuminemia, de ned as serum albumin < 35 g/l; (3) elevated lactate dehydrogenase

Follow-up
Enrolled patients were followed until January 2020. Data were collected through telephone, letters, and case records. Survival time was de ned as the period from diagnosis to death or last interview. None of the enrolled patients were lost to follow-up.

Statistical analysis
Data were analyzed with SPSS 26.0 software for windows (SPSS, Inc., Chicago, IL, USA). The c 2 -test was used to analyze the relationship between pathological/clinical subtypes and clinical features. The Kaplan-Meier method was applied to analyze the survival curve. The Log-Rank test was used to compare the differences in the survival curve. A Cox regression model was employed to perform univariate analysis and multivariate analysis in order to identify independent prognostic factors for survival. A Pvalue < 0.05 was considered to be statistically signi cant.

Patient characteristics
All 54 patients diagnosed with CD were hospitalized between February 2008 and August 2018. Within this cohort, 23 patients were aged from 12 to 40 years old, and 31 patients were between 41 and 82 years old (median, 43 years). A total of 24 patients were male and 30 were female. Based on the clinical subtype, 30 patients had UCD and 24 had MCD. According to the histopathological characteristics of all 54 specimens, 32 cases were HV, 19 cases were PC, and only 3 cases were Mix ( Table 1).
The main complaints of the 54 CD patients were categorized into four groups, and each patient may have had one or more complaints. The rst group included 31 patients with enlarged super cial lymph nodes or serendipitous tumor masses, which were con rmed to be CD after biopsies or surgeries. The second group included 12 patients with recurring symptoms, such as fever, hypodynamia, or myalgia. After physical examination, ultrasonography, or CT, enlarged lymph nodes or tumor masses were found and then con rmed to be CD after biopsies or surgeries. The third group included 9 patients with skin ulcers, blisters, or stomatitis, which were considered to be PNP and were diagnosed as CD after biopsies or surgeries. The remaining 10 cases complained of non-typical symptoms, such as abdominal distension, pain, or body edema (Table 1).

Clinical symptoms and complications
At the time of hospitalization, several obvious signs and symptoms occurred in CD patients, including fever, splenomegaly, and pleural effusion and/or ascites. Fever was observed in 11 patients and splenomegaly was observed in 12 patients. Pleural effusion and/or ascites were determined in 16 patients by CT scans (Table 2).
Pulmonary infection was diagnosed in 16 patients whose main complaints were fever and cough, and Xray lm or CT assisted with the de nite diagnosis. A total of 18 cases were diagnosed with kidney injury, based on proteinuria and signi cantly elevated serum creatinine. Liver injury was observed in 9 patients with abnormal elevation of serum alanine aminotransferase (ALT) and glutamic oxalacetic transaminase (AST). In total, 18 patents were diagnosed with anemia, according to obviously decreased hemoglobin, and 3 patients were diagnosed with autoimmune hemolytic anemia (AIHA) by positive Coombs' test results. Nine patients with skin involvement were diagnosed with PNP, with chief complaints of skin or mucosal ulcers, blisters, or pigmentation. Six MCD cases were diagnosed with POEMS syndrome, with presence of polyneuropathy, organomegaly, endocrinopathy, monoclonal gammopathy, and skin changes ( Table 2).

Patients with MCD may exhibit more symptoms and complications
In this retrospective study, 30 (55.6%) patients were diagnosed with UCD and 24 (44.4%) patients were diagnosed with MCD (Table 1). MCD patients commonly exhibited clinical signs and symptoms, including in 9 of the 11 patients with fever (P < 0.01), 9 of the 12 patients with splenomegaly (P < 0.05), and 11 of the 16 patients with pleural effusion and/or ascites (P < 0.05) ( Table 2).
In all CD patients with clinical complications, liver injury, anemia, and POEMS syndrome were more likely to be found in MCD patients. Of the 9 patients with liver injury, 6 had MCD, and of the 19 patients with anemia, 11 had MCD; however, these differences were not statistically signi cant (P = 0.165 and P = 0.143, respectively). All 6 patients with POEMS syndrome were diagnosed with MCD (P < 0.01). There were no signi cant differences in the remaining clinical complications between patients with UCD and MCD. Furthermore, the MCD subtype was found in 7 of the 16 patients with pulmonary infection, in 10 of the 18 patients with kidney injury, in 1 of the 3 patients with AIHA, and in 5 of the 9 patients with PNP ( Table 2).

Relationship between pathological subtypes and clinical symptoms and complications
Of the total included patients, 32 (59.3%) cases were HV, 19 (35.2%) cases were PC, and only 3 (5.6%) cases were Mix (Table 1). Since the sample size of the Mix cases was too small for statistical analysis, only PC and HV cases were used for analysis. In terms of the signs and symptoms, fever was more likely to occur in patients with PC than in patients with HV (P < 0.05). With regards to clinical complications, all 3 patients with AIHA were diagnosed with PC (P < 0.05), and of the 9 patients with PNP, 5 patients were classi ed as PC and 3 were classi ed as HV; however, this difference was not statistically signi cant (P = 0.131) ( Table 2).

Patients with MCD may show abnormal laboratory parameters
Of the 54 included patients, 17 showed a pretreatment albumin level of < 35 g/l. Patients with PC were more likely to have lower serum albumin than those with HV (P < 0.05). Prior to treatment, 6 patients showed elevated lactate dehydrogenase (LDH) levels, which tended to be more common in patients with MCD or PC, although no signi cant difference was found (P = 0.078, P = 0.058). Prior to treatment, patients with MCD also showed elevated CRP and ESR levels; of the 10 patients with elevated CRP, 9 were diagnosed with MCD (P < 0.01), and of the 13 patients with elevated ESR, 11 were diagnosed with MCD (P < 0.01). Eight patients had an ASO level > 200 IU/ml, and patients with PC were more likely to have elevated ASO (P < 0.01) ( Table 2).

Treatment
Two patients who complained of enlarged super cial lymph nodes only received lymph node biopsy, and both refused further treatment. A total of 23 patients without serious complications received surgery, and then took a "watch and wait" strategy. Eight patients were treated with the CHOP regimen (cyclophosphamide 600 mg/m 2 , vincristine 1 mg/m 2 , and prednisone 1 mg/kg) after surgery, and 7 were classi ed as UCD. In MCD cases, 9 received CHOP, 6 received R-CHOP (at least two doses of rituximab 375 mg/m 2 ). Patients with PNP received standard treatment with intravenous infusion of immunoglobulin (IVIG) and glucocorticoid (prednisone, methylprednisolone, or dexamethasone). Two patients in critical condition received supportive treatment only, including anti-infection, blood pressure control, and hemodialysis, which led to symptomatic improvement (Table 1).
Univariate analysis identi ed PNP and elevated CRP as unfavorable risk factors Among the 54 evaluated cases, the longest follow-up duration was 143 months, and the median followup duration was 57.5 months. Univariate analysis of prognostic factors using the Cox univariate analysis identi ed two risk factors: Presence of PNP (HR = 31.895, P < 0.01) and elevated CRP (HR = 5.363, P < 0.05) ( Table 3). Kaplan-Meier analysis and log-rank test also indicated a signi cantly shorter survival for patients with PNP (P < 0.001) or elevated CRP (P = 0.021) (Fig 2). In addition, univariate analysis showed that fever, pleural effusion and/or ascites, and low serum albumin level may be unfavorable risk factors, but these results were not statistically signi cant (0.05 < P < 0.1) ( Table 3).

Multivariate analysis identi ed PNP as the only risk factor
A Cox proportional hazards model was used for multivariate analysis, in which characteristics with Pvalues < 0.15 in univariate analysis and those with clinical signi cance, were included. These characteristics included fever, pleural effusion and/or ascites, PNP, low serum albumin, and elevated CRP. Multivariate analysis showed that the presence of PNP was as independent risk factor associated with the prognosis of CD (HR = 22.834, P < 0.01). Although elevated CRP was identi ed as an unfavorable risk factor in univariate analysis, its P-value was 0.639 in multivariate analysis (Table 3).

Discussion
Compared to other common hemopathy, such as leukemia and lymphoma, research on CD is limited due to its rarity. Although CD is not a malignant disease, it has been associated with an increased risk of diverse complications. Although studies have focused on establishing the diagnostic criteria of CD [13,14], additional cohort studies are required to better understand its prognosis factors.
Our study retrospectively analyzed 54 patients with CD in a single center from 2008 to 2018. We found that MCD may present with more systemic manifestations such as fever, splenomegaly, and pleural effusion and/or ascites. Furthermore, POEMS syndrome, as a complication, also occurred more frequently in patients with MCD. MCD commonly presented with serological abnormalities corresponding to in ammatory markers, including elevated CRP and ESR levels. These results indicated that MCD may induce systemic in ammation; thus, systemic therapies were primarily applied to MCD. In terms of pathology, PC cases were more likely to have fever, and decreased albumin and elevated ASO were often detected in the serum of PC patients.
Univariate analysis identi ed the presence of PNP and elevated CRP in serum as risk factors in uencing the survival of CD patients. However, when all candidate risk factors, including fever, pleural effusion and/or ascites, PNP, low serum albumin, and elevated CRP were included in multivariate analysis, the presence of PNP was the only independent unfavorable risk factor for the prognosis of CD. This result was consistent with the research of Dr. Dong [3].
PNP is a rare mucocutaneous autoimmune disease associated with neoplasm that was rst described in 1990 [15]. The clinical features of PNP include stomatitis, mucositis, and skin lesions. Furthermore, PNP is frequently associated with hematologic neoplasms, including non-Hodgkin lymphoma, chronic lymphocytic leukemia, and CD [16,17]. In this series of CD cases, 9 were considered to have PNP. Clinical classi cation showed that 4 were UCD and 5 were MCD, while pathological classi cation showed that 5 were PC, 1 was Mix, and 3 were HV. Although several previous studies have reported that PNP usually occurs in UCD or HV [3,15], our study found no correlation between PNP and clinical or pathological subtype. The main complaints of these patients were polymorphic skin lesions, including skin blisters, ulcers, and lichenoid eruptions, while stomatitis and mucositis were also observed. Some studies revealed that patients with PNP tended to die from severe infection due to immunosuppressive therapy, associated malignancy, and bronchiolitis obliterans [18]. Thus, with poor prognosis and high mortality, the treatment of PNP is challenging. Patients with PNP should receive systemic corticosteroids combined with other immunosuppressive agents, including cyclosporine, cyclophosphamide, azathioprine, and mycophenolate mofetil [19]. In our retrospective study, all 9 CD patients with PNP received IVIG and steroids, but 5 died by the date of the last follow-up.
The centricity and pathology type are important clinical factors that could help to predict prognosis and guide treatment at early diagnosis. Several recent studies have reported that MCD patients have signi cantly worse survival rates than UCD patients [20,21]. Furthermore, based on pathological classi cation, PC patients were also reported to have a worse prognosis than both HV and Mix patients [13]. Unfortunately univariate analysis in our study could not identify centricity (UCD or MCD) and histopathology types (HV or PC) as prognostic factors in this series of patients. It will be important to collect more cases of CD for further analysis in order to investigate the correlations between the centricity/pathology type and the prognosis.
Complete resection of the tumor mass was reported to be the standard treatment for UCD [22]. Among 30 UCD patients, 20 cases only received surgical resection, 7 cases received the CHOP regimen after surgery, 1 case was too sick to tolerate surgery and only received symptomatic and supportive treatment. For MCD, the optimal treatment for has not been well established. The MCD patients in our study received a variety of agents, including corticosteroids, cytotoxic chemotherapy, immunoglobulin, rituximab, and anti-IL-6 (tocilizumab). MCD patients could bene t from cytotoxic chemotherapy based on that used in lymphoma therapy [23]. In our study, most of MCD the patients received cytotoxic chemotherapy as a rst line therapy. Rituximab, a monoclonal anti-CD20 antibody, was used in HIV-and/or HHV8-positive MCD patients [24,25]. Tocilizumab, a humanized anti-IL-6 monoclonal antibody was approved for treatment of CD in Japan in 2005[26], and has been shown to induce remission in MCD patients in a series of case reports[27, 28]. These target therapy regimens have the potential to be alternative treatments for CD after replacement of chemotherapy. Due to the high heterogeneity of CD, precision and individual therapy should be urgently applied in the clinic.
The present study had some limitations. First, it was a retrospective study and there might be a bias for patient selection and data collection. Second, the sample size requires to be expanded for further analysis.

Conclusions
CD was an unusual lymphoproliferative disorder that continues to present clinical challenges. Our study helped to identify the clinical characteristics and prognosis of CD patients. The results indicated that the presence of PNP was an independent risk factor, and should be paid more attention during diagnosis and treatment.

Funding
The study was supported by National Natural Science Foundation of China (81800587). Clinical follow up and data analysis were supported by this funding.

Availability of data and materials
All methods were carried out in accordance with relevant guidelines and regulations. The data and materials are available.

Ethics Declarations
Ethics approval and consent to participate Written Informed consent was obtained from all the adult patients and parents or guardians for participants under 18 years old. Dead patients' kin or legally authorized representative provided written informed consent. This study was approved by ethical committee of The First A liated Hospital of Nanjing Medical University.

Consent for publication
Written informed consent for publication was obtained from all participants.