Improving the Diagnosis and Assessment of Castleman Disease

Castleman disease (CD) is a rare, heterogeneous, cytokine-driven disorder with unicentric and multicentric subtypes (i.e., involving one or multiple lymph node regions, respectively). Its pathogenesis is poorly understood, and the CD diagnostic process is often one of exclusion. 

The diagnostic difficulty is compounded by a myriad of factors, including overlap with many infectious, autoimmune, and oncologic conditions whose symptoms may mimic those of CD. Consequently, patients commonly experience significant delays in the diagnosis of CD, ranging from months to years, and some patients undergo a challenging clinical journey fraught with multiple misdiagnoses. 

Moreover, interpretation of lymph node histology is quite complex, with significant inter-observer variability, further complicating diagnosis and blurring prognostic clarity.

The inherent challenges of diagnosing CD, including grading of the lymph node biopsy, have prompted Recordati to collaborate with the ARUP Institute for Research & Innovation on initiatives designed to facilitate earlier and more consistent diagnosis of CD. 

Automated grading of Castleman disease histopathology

One initiative involves the development of the first proof-of-concept machine learning model for recognizing and grading Castleman-like histology features. 

At the recent 67th American Society of Hematology (ASH) annual meeting, researchers demonstrated the feasibility of using the attention-based, multiple learning, artificial intelligence (AI) model, the predictive power of which compared favorably to expert hemopathologist consensus.¹ 

Continued development of the model will focus on optimizing the model’s performance through acquisition of more diverse data, ultimately enabling faster, more accurate, and more consistent diagnosis of CD.

The real-world impact of idiopathic multicentric Castleman disease

Additionally, ARUP is one of several institutions collaborating with Recordati on the BURDEN-iMCD study, which is designed to characterize the economic impact of Castleman disease and related morbidities in patients with the idiopathic multicentric subtype of CD (iMCD), so-called because of its unknown etiology.² 

Characterized by diffuse lymphadenopathy and systemic inflammation, iMCD accounts for one-third to one-half of all CD cases but has no validated diagnostic biomarkers, making diagnosis especially challenging. 

Further, there is limited information about the impact of iMCD on both burden of illness (i.e., comorbidities) and healthcare resource utilization (HCRU). Notably, previous analyses of datasets that included patients with iMCD were based on nonspecific ICD codes (e.g., lymph node enlargement) and likely included other subtypes of CD such as the unicentric variant (UCD), for which prognosis differs from that of individuals with iMCD.

A retrospective, real-world analysis of iMCD administrative claims data from the Merative™ MarketScan® Databases from January 1, 2016, through June 30, 2024, the BURDEN-iMCD study is evaluating comorbidities and HCRU in the six months following an iMCD diagnosis, using a refined patient identification algorithm. 

As recently reported at the ASH annual meeting, the study documented a tremendous burden of disease and significantly increased HCRU in 140 patients with iMCD compared to 420 non-iMCD controls. The prevalence of comorbidities such as anemia, renal dysfunction, and respiratory dysfunction or interstitial lung disease was 6.9-fold (95 percent confidence interval [CI]: 6.7, 7.2) greater among patients with iMCD than in controls. The observed morbidities resulted in associated healthcare costs that were 7.61-fold higher in the iMCD patient cohort than in controls.

It should be noted that the data from the BURDEN-iMCD analysis are retrospective and have less evidentiary value than prospective studies. Additionally, the analysis does not establish causality. Nevertheless, the BURDEN-iMCD results have important implications for identifying and managing patients with iMCD, in that earlier diagnosis and initiation of evidence-based treatment would presumably help patients avoid worsening morbidities and complications associated with the disease.

We look forward to ongoing and future research initiatives that shed additional light on the real-world impact of iMCD on patients’ lives, and that such findings motivate researchers and practitioners to step up efforts to facilitate earlier recognition and optimized management of CD.

References:

1. Morrison M, O’Fallon B, Hutchings, A, et al. Automated grading of Castleman disease histopathology using an attention-based multiple-instance learning model. Presented at 67th American Society of Hematology (ASH) Annual Meeting and Exposition, Orlando, Fla., December 7, 2025. Poster #3002.

2. Noy A, Ohgami R, Munshi N, et al, Retrospective real-world data analysis of morbidity burden and healthcare costs in idiopathic multicentric Castleman disease compared with matched controls (BURDEN-iMCD). Presented at 67th American Society of Hematology (ASH) Annual Meeting and Exposition, Orlando, Fla., December 6, 2025. Poster #2606.