Clinical features and treatment outcomes of <scp>POEMS</scp> ‐associated Castleman disease

POEMS syndrome (Polyneuropathy, Organomegaly, Endocrinopathy, Monoclonal Protein, Skin Changes) is a rare multisystem disorder characterised by plasma cell dyscrasia and a typical neuropathy. Castleman disease is one of the five major diagnostic criteria for POEMS syndrome, with published cohorts reporting its coexistence in 11%–30% of cases.1-3 The rarity of POEMS-associated Castleman disease means limited data on its clinical characteristics and treatment outcomes are available. A study by Li et al. reported no significant differences in the clinical characteristics of POEMS patients with (POEMS-CD) and without (POEMS) Castleman disease, except for a higher frequency of thrombocytosis in patients with POEMS-CD.2 Researchers from the Mayo Clinic found that the peripheral neuropathy in POEMS-CD exhibited a similar but less severe phenotype compared to the POEMS cohort.4 Most reported POEMS-CD cases are associated with multicentric Castleman disease (MCD), whereas the potential occurrence of unicentric Castleman disease (UCD) in this setting remains unclear. Data on the temporal relationship between Castleman disease and POEMS syndrome are also lacking. Given that previously reported cases of POEMS-CD are limited, this study aims to add to the existing literature by comparing the clinical features and treatment outcomes of patients with POEMS-CD and POEMS, treated at University College London Hospital (UCLH), a national referral centre in the United Kingdom. A retrospective review of all patients with an established diagnosis of POEMS syndrome referred to the national POEMS service, a multidisciplinary neuro-haematology service, at UCLH, from 1998 to August 2024 was performed. This study included patients with POEMS syndrome who also had a histological diagnosis of Castleman disease as well as those with Castleman disease variant of POEMS. Patients with Castleman-like histology were excluded. Computed tomography (CT) or positron emission tomography/computed tomography (PET/CT) was performed to establish the extent of nodal involvement and to differentiate UCD from MCD. Clinical assessments along with laboratory, radiological, histopathological findings and treatment details were retrieved. Treatment responses were assessed in accordance with published guidelines.5 Details of neurological assessment, response evaluation criteria and statistical methods are outlined in Supporting Information. This study included 150 patients with POEMS syndrome diagnosed between 1998 and 2024. Castleman disease was confirmed histologically in 18 (12%) cases. The clinical characteristics of patients with POEMS-CD and POEMS are summarised and compared in Table S1. The median age at presentation was significantly younger in the POEMS-CD group (45 years) than in the POEMS group (55 years). The median time to diagnosis was similar in both groups, being 8.5 months (interquartile range [IQR] 5–27 months) for patients with POEMS-CD and 7 months (IQR 7–21 months) for those with POEMS only. The median paraprotein at diagnosis was 2 g/L for both groups. The median VEGF level at presentation was higher in patients with POEMS-CD (5690 pg/mL) than in those with POEMS (3869 pg/mL), although this difference did not reach statistical significance. No statistically significant differences were observed between the two groups regarding gender, endocrinopathy, extravascular volume overload, sclerotic bone lesions, polycythaemia, thrombocytosis or skin changes. Neuropathy was present in 83% (15/18) of the POEMS-CD group and 95% (125/132) of the POEMS group. Based on neurophysiology, almost all patients had a sensorimotor neuropathy involvement, with 93% (13/14) in the POEMS-CD group and 94% (102/109) in the POEMS group. Conduction slowing, often referred to as ‘demyelination,’ was noted in 79% (11/14) of the POEMS-CD group and 87% (95/109) of the POEMS group. The median Rasch-built overall disability scale (RODS) score was 57 for the POEMS-CD group and 48 for the POEMS group (higher percentile RODS better), though this difference was not statistically significant (Table 1). The median neuropathy impairment score (NIS) was significantly lower for the POEMS-CD group (27.5) than the POEMS group (62.9), with lower scores indicating lesser neuropathy severity. Similarly, the neurophysiological variables both showed significantly higher median composite amplitude score (CAS) and compound muscle action potential (CMAP) amplitude of tibialis anterior in the POEMS-CD group than the POEMS group (higher scores correspond to lower neuropathy severity). Overall, these findings indicate that patients with POEMS-CD had a milder neuropathy than those with POEMS only. In most cases (77%), the diagnoses of POEMS syndrome and Castleman disease were made either concurrently or in close succession, with all of these patients having MCD. However, three patients were diagnosed with Castleman disease years prior to the onset of other symptoms of POEMS syndrome, with intervals ranging from 68 to 114 months. Two had UCD and underwent surgical excision; one with MCD was treated with systemic chemotherapy before developing POEMS 102 months later. The hyaline vascular (HV) variant was the most commonly reported histological subtype, present in 56% of patients, followed by the plasma cell (PC) variant (28%) and the mixed variant (17%) with histological features of both HV and PC variants. None of these patients had follicular dendritic cell sarcoma. We evaluated the treatment outcomes of 60 patients who underwent autologous stem cell transplantation (ASCT) between 1998 and 2024, including nine diagnosed with Castleman disease. The first-line treatment for these patients can be found in Table S2. Among patients with POEMS-CD, 78% received induction chemotherapy prior to ASCT, while 22% underwent upfront ASCT. In comparison, for those with POEMS only, 53% received induction chemotherapy followed by ASCT and 47% were treated with upfront ASCT. The most commonly prescribed induction regimen was lenalidomide and dexamethasone. The treatment responses are summarised in Table S3. While the VEGF overall response rates (ORRs) were over 90% in both groups, a lower haematological ORR was noted in those with POEMS-CD (67% vs. 78%). In patients with POEMS-CD, the rate of CRVEGF was 89% and CRH was 33%, compared to 80% and 57% in those with POEMS. The rates of relapsed or refractory disease among patients with POEMS and POEMS-CD were 33% and 44% respectively. The median follow-up period for the entire cohort was 126 months. The 5-year PFS was 38% for the POEMS-CD cohort and 74% for the POEMS cohort. The median PFS was 49 months for the POEMS-CD group compared to 126 months for the POEMS group, although this difference did not reach statistical significance (p = 0.228). These findings suggest a trend towards shorter PFS in POEMS-CD (HR 1.90; 95% CI, 0.63–5.72; p = 0.25) (Figure 1A). This trend persists despite the use of induction therapy prior to ASCT (Figure 1B) and the administration of lenalidomide-containing induction (Figure 1C). This study represents one of the largest single-institution analyses of POEMS-CD to date and provides comprehensive data on the clinical presentation, neurological profiling and outcomes of this rare entity. Our study did not reveal significant differences in the clinical features between the two cohorts although it is limited by small numbers. The presence of Castleman disease may not substantially alter the overall broader clinical presentation of POEMS syndrome. While we identified very frequent neuropathy, largely with a prominent conduction slowing (demyelinating) sensorimotor phenotype, the severity of the peripheral neuropathy was significantly less in those with POEMS-CD than those with POEMS syndrome alone. This finding aligns with observations from the Mayo researchers.4 The exact pathophysiology underlying this phenomenon remains unclear. Both VEGF and interleukin-6 (IL-6) play crucial roles in stimulating angiogenesis within the tumour microenvironment. Unlike VEGF, however, IL-6-driven angiogenesis has been linked to defective pericyte coverage, as seen in animal studies.6 Pericytes are crucial for formation and regulation of the blood-nerve barrier (BNB).7 Hypercytokinaemia, particularly high levels of VEGF and IL-6, significantly contributes to increased vascular permeability, resulting in complications such as peripheral oedema, ascites, pleural effusion and papilledema. Pericyte dysfunction can disrupt the integrity of the BNB, leading to neuropathy through endoneurial oedema and exposing the nerves to neurotoxic cytokines. Sural nerve biopsies in patients with POEMS syndrome show oedema, neovascularization and endothelial hyperplasia, likely reflecting some impact of IL-6 and VEGF on the BNB.8, 9 Notably, several studies have failed to demonstrate elevated IL-6 levels in the serum of some patients with Castleman disease.10-12 It is hypothesized that the consistently high levels of IL-6 and VEGF in POEMS syndrome patients may cause greater endothelial disruption, leading to vascular leakage and potentially neuropathy. Investigating the cytokine profiles among Castleman disease, POEMS-CD and POEMS alone may elucidate the mechanisms underlying the varying severity of neuropathy observed. While the VEGF response rates were comparable in both groups, patients with POEMS-CD experienced lower haematological response rates. A trend towards a shorter PFS was observed among patients with POEMS-CD, although the difference did not reach statistical significance, likely due to small sample size. Current treatments for POEMS-CD, adapted from the myeloma armamentarium, focus on eradication of the monoclonal plasma cells that drive the disease process. Most patients in our cohort did not receive treatment specific to Castleman disease. Future studies might evaluate any role of adjunctive anti-IL-6 in POEMS-CD. Interestingly, our data suggest that Castleman disease can precede the development of POEMS syndrome by several years, highlighting the need for vigilance for POEMS symptoms in patients with Castleman disease. Both UCD and MCD can be associated with POEMS syndrome. However, given that only two cases of UCD were observed, it is not possible to draw definitive conclusions regarding potential differences between UCD and MCD in POEMS-CD. Our study has several limitations. The retrospective nature of our study and small sample size may limit the robustness of our conclusions. A centralised review of all the histological specimens was not conducted, which represents another limitation. The heterogeneity of induction regimens across our cohort may affect the comparability of treatment outcomes. Future studies should compare the cytokine profiles among CD, POEMS-CD and POEMS and explore the role of anti-IL-6 in POEMS-CD, focusing on its impact on treatment response and survival outcomes. G.L., V.J.W.Z. and Z.M.-O. collected data, performed analyses and wrote the draft; J.S. and M.P.L. supervised the study; and all authors critically appraised the study and contributed to the final manuscript. M.P.L. is supported by the UCLH NHS Biomedical Research Centre. The authors declare that no specific funding was received for this research. S.H.G. has received honoraria, speaker fees and travel support from Recordati. The remaining authors declare no competing financial interests. The study was performed using anonymised data and was conducted in accordance with the Declaration of Helsinski. The Research Ethics Committee has waived the requirement for formal ethics approval due to the absence of identifiable information. Written informed consent was waived due to the retrospective study design. The datasets analysed for this study are available from the corresponding author upon reasonable request. Data S1. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.