A total of 276 treatment-naïve patients with Castleman Disease (CD), diagnosed at our institution between November 2013 and April 2023, were identified through the information system of The First Affiliated Hospital of Zhengzhou University.

The initial classification yielded 81 cases of UCD, 39 cases of OCD, and 156 cases of iMCD. In this study, all cases underwent a central review by two expert pathologists upon patient enrolment to ensure the exclusion of cases not meeting the classic diagnostic criteria for CD. We ultimately identified a final cohort consisting of 74(29.8%) cases of UCD, 38(15.3%) cases of OCD, and 136(54.8%) cases of iMCD. The present study focuses on the UCD and OCD subtypes due to their greater clinical similarity. The study was approved by the Ethics Committee of the First Affiliated Hospital of Zhengzhou University (Ethical Approval No. 2022-KY-0869-001).

The diagnoses of UCD and OCD were re-confirmed by reviewing admission records, imaging reports, and laboratory tests. The imaging information of the cases was re-evaluated to determine whether the disease was unicentric or oligocentric. Cases with concurrent tumours, infections, or autoimmune diseases were excluded from both groups, as these conditions might accompany CD lymph node pathological changes. Ultimately, 63 cases were included in the UCD group and 37 in the OCD group (Figures 1a, b). Demographic, clinical, laboratory, and treatment-related data, including age, gender, presenting symptoms, physical examination findings, imaging and laboratory tests, histopathology, treatment strategies, and outcomes, were extracted from the patients’ medical records.

In the absence of standardised response criteria for OCD, our efficacy assessment was based on the established framework for iMCD, which includes laboratory results: C-reactive protein (CRP), hemoglobin (HGB), albumin, glomerular filtration rate (GFR), systemic symptoms (fatigue, anorexia, fever, weight loss), and the largest diameter of enlarged lymph nodes.① Complete Response (CR) is defined as normal laboratory test results and systemic symptoms returning to normal.② Partial Response (PR) is defined as improvement in 4 systemic symptoms, laboratory results improving by more than 50%, and a reduction of more than 50% in the largest diameter of the enlarged lymph nodes.③ Progressive Disease (PD) is defined as a worsening of laboratory test results by more than 25%, any of the 4 systemic symptoms related to inflammation worsening, an increase of more than 25% in the largest diameter of enlarged lymph nodes, or relapse.④ Stable Disease (SD) is defined as the absence of partial or complete remission or progressive disease (16).

Follow-up was conducted by telephone, reviewing patients’ inpatient records, and outpatient follow-up records, with the follow-up cutoff date of May 31, 2025. The median follow-up time was 67.0 (25-141) months. The primary endpoints for this study were Overall Survival (OS) and Progression-Free Survival (PFS).

Descriptive statistics were used to summarise case characteristics. Clinical data were compared between the UCD and OligoCD groups using the chi-squared test, corrected chi-squared test, Fisher’s exact test, independent samples t-test, or Mann-Whitney U test, as appropriate. OS and PFS was estimated using the Kaplan-Meier method. Univariate analysis for PFS involved group comparisons using the log-rank test and calculation of Hazard Ratios (HRs) with 95% Confidence Intervals (CIs) using univariate Cox proportional hazards models. A P-value of less than 0.05 was considered statistically significant. All statistical analyses were performed using SPSS version 25.0 (IBM, Armonk, NY) and R software version 4.5.1.

There is no significant difference in gender or age distribution was observed between the two groups. In the UCD group, the maximum diameter of the enlarged lymph nodes (MDLN) was 5.55 ± 2.46 cm, and the lymph node region affected by the largest lymph node (LNRALN) was most commonly found in the abdomen/retroperitoneum (36.5%). In the OCD group, the MDLN was 5.14 ± 1.66 cm. The LNRALN was most commonly found in the neck (48.6%). There was a significant difference in LNRALN between the two groups (p=0.021) (Table 1, Figure 2). All patients presented with disease on one side of the diaphragm.

There was a significant difference in pathological classification between the two groups (p=0.003), with the HV-CDtype being more prevalent in the UCD group (90.5% vs 62.2%) (p=0.001); the mixed type was more prevalent in the aMCD group (27.8% vs 6.3%) (p=0.004) (Table 1).

We compiled the reasons for patients’ hospital admission and consultation, defining systemic symptoms. The reasons for presentation did not differ significantly between the two groups; in both cohorts, the majority of cases were diagnosed incidentally(74.6% vs 75.7%), while patients presenting with systemic symptoms represented a minority(4.8% vs 5.4%). (Table 1) Compared with the UCD group, the OCD group demonstrated a significantly higher proportion of cases with elevated levels of ESR and CRP (26.7% vs 5.7%, 30.0% vs 3.8%, respectively), with all differences being statistically significant (Table 2).

In the OCD cohort, 23(63.9%) cases involved only two lymph node regions, while 14 cases (36.1%) involved ≥3 lymph node regions.

To assess the internal heterogeneity of disease activity, 9 (24.3%) patients underwent 18F-Fluorodeoxyglucose Positron Emission Tomography/Computed Tomography (PET-CT) scans. The difference in Standard Uptake Value Maximum (SUVmax) between the largest lymph node and the other lymph node with the highest metabolic activity was not statistically significant (6.40 vs. 3.94) (t=1.836, P = 0.104).

Furthermore, to assess the heterogeneity of tumour bulk, a paired-samples t-test was performed comparing the Maximum Diameter of the Enlarged Lymph Nodes (MDLN) to the diameter of the second-largest involved lymph node in the OCD cohort (N = 37). The analysis revealed a highly statistically significant difference in size between the two lesion groups (MDLN, 5.14 ± 1.66cm vs. second-largest, 1.76 ± 0.44cm; t=13.281, P<0.001).

The median follow-up time was 67.0 (25-141) months. At the end of the follow-up period, no deaths occurred, except in cases lost to follow-up. Consequently, the primary analytical focus shifts to PFS as a more clinically relevant endpoint for assessing disease-related morbidity and control.

We performed a univariate analysis using the Kaplan-Meier method and the log-rank test to identify factors at diagnosis that were predictive of PFS. The results, revealed several prognostic indicators. Histopathological subtype was a key determinant of outcome. While the Mixed-CD type was not significantly associated with a poorer PFS (HR 2.62; 95% CI 0.37–18.6; P = 0.3), PC-CD had a profound and statistically significant negative impact on PFS (HR, 17.6; 95% CI 2.78–111; P = 0.002) (Table 3, Figure 3f).

Factors related to the anatomical burden and distribution of disease were also associated with progression. Specifically, non-contiguous nodal involvement(COINR Non-contiguous) (HR 20.9; 95% CI 3.92–112; P < 0.001), and the involvement of three or more nodal regions(NOIOR≥3) (HR 11.4; 95% CI 1.37–95.3; P = 0.024) were all highly significant predictors of a shorter PFS (Table 3, Figure 3e).

Furthermore, a hyper-inflammatory state at diagnosis, manifested by elevated systemic inflammatory markers, was also significantly associated with a poor prognosis. This was evidenced by elevated levels of CRP (>10.00 mg/L: HR 14.3; 95% CI 1.65–123; P = 0.016) and ESR (>15 mm/h: HR 17.6; 95% CI 2.03–152; P = 0.009). Finally, failure to achieve a CR to initial therapy was also a significant predictor of subsequent progression (HR 4.7; 95% CI 1.05–21.0; P = 0.043) (Table 3, Figures 3c, d, g).

We attempted to construct a multivariate Cox proportional hazards model to identify independent prognostic factors. However, due to the small number of endpoint events in this cohort (only 6 progression events among 37 patients), the model was statistically unstable and demonstrated the phenomenon of complete separation. This issue precluded the calculation of valid hazard ratios, rendering the multivariate analysis uninformative.

To contextualise our findings within the evolving nosology of Castleman disease, we compared the clinicopathological features and outcomes of our OCD cohort with two recently described and highly relevant patient populations: the initial OCD cohort described by Pierson et al. (13), and the asymptomatic multicentric Castleman disease (aMCD) cohort reported by Zhang et al. (Table 4).

Our study cohort (N = 37) is substantially larger than the seminal cohort (N = 14) originally used by Pierson et al. to define the OligoCD entity. While key demographic features such as mean age were comparable (35.4 vs. 34.0 years), critical differences exist in disease distribution and pathology. Consistent with the cohort described by Pierson and colleagues, our patients also presented with disease exclusively confined to one side of the diaphragm. A critical difference, however, was observed in the histopathology. Our cohort included a significant proportion of cases with the PC-CD subtype (10.8%), whereas this variant was entirely absent from the Pierson cohort (0.0%).

A comparison with the large, multicentric cohort of 114 aMCD patients described by Zhang et al. provides a robust benchmark for indolent yet multicentric disease. Despite these differing management philosophies, the most striking finding was the remarkably similar rate of progression to iMCD. In our cohort, 2 of 37 cases (5.4%) progressed to iMCD during follow-up, a proportion nearly identical to that in the cohort by Zhang et al., where 6 of 114 cases (5.3%) underwent a similar transformation.

Given the significant heterogeneity observed in PFS during the univariate analysis, we conducted an exploratory subgroup analysis to stratify the OCD cohort into more clinically and biologically homogeneous subgroups. Cases were categorised into either an ‘asymptomatic group’, defined by the absence of systemic symptoms and normal inflammatory markers (analogous to the aMCD entity), or a ‘high-inflammatory group’, which included cases with either abnormal laboratory parameters, primarily elevated inflammatory markers. or systemic symptoms that did not meet the formal diagnostic criteria for iMCD. The core finding of this analysis was the profound and highly statistically significant difference in PFS between these two subgroups. Kaplan-Meier analysis revealed that patients in the high-inflammatory group had a markedly poorer PFS compared with those in the asymptomatic group (log-rank test, P = 0.042) (Figure 3b).

Subsequent analysis of the clinicopathological characteristics of these subgroups revealed a distinct set of high-risk features defining the ‘high-inflammatory’ phenotype. A significant difference was observed in the distribution of histopathological subtypes between the two groups (overall P = 0.018). This difference was primarily driven by a significantly lower prevalence of the indolent HV-CD subtype in the high-inflammatory group (P = 0.017). Notably, all four PC-CD cases in the cohort were found within the high-inflammatory group, although this specific comparison did not reach statistical significance (P = 0.105) (Table 5).

Significant differences were also noted in the pattern of nodal involvement. The high-inflammatory group was characterised by a more widespread and systemic distribution of disease, with a significantly higher rate of non-contiguous nodal involvement (P = 0.008) and a greater number of involved nodal regions (P = 0.017) (Table 5).