Efficacy and safety of steroid‐sparing strategy in the treatment of AIHA associated with KSHV/HHV‐8 positive multicentric Castleman disease

Multicentric Castleman disease (MCD) related to HHV-8 (human herpesvirus-8 or Kaposi sarcoma-associated herpesvirus, KSHV) is a lymphoproliferative disorder mainly affecting immunocompromised individuals. Flares present with lymphadenopathy, splenomegaly, inflammation, and high HHV-8 viremia.1, 2 Diagnosis requires a lymph node biopsy showing plasma-cell or mixed type MCD features with infected large B cells in the mantle zone. Life-threatening complications include hemophagocytic lympho-histiocytosis (HLH) and HHV-8-related lymphoma, while autoimmune manifestations such as dysglycemia, thrombotic thrombocytopenic purpura (TTP), and cytopenias are also observed.2-5 Standard of care relies on rituximab, often combined with etoposide or liposomal doxorubicin, for the management of acute flares.6, 7 Autoimmune hemolytic anemia (AIHA) is an acquired autoimmune disorder caused by autoantibody-mediated red blood cells (RBCs) destruction. It is classified as warm or cold AIHA, according to the antibodies' thermal range. Corticosteroids (CS) remain the first-line therapy for primary warm AIHA (wAIHA), with rituximab used as a second-line treatment.8 HHV-8 MCD is a rare but recognized cause of wAIHA, as documented in a few case reports, but the characteristics, evolution, and management of this complication have never been fully addressed. CS are often used with rituximab, as in primary wAIHA, but this strategy might increase Kaposi sarcoma (KS) risk and worsen immune deficiency.9, 10 Patients were identified through the French National Reference Center for Castleman Disease (see details in Supplemental Methods). AIHA was defined by hemoglobin (Hb) 12 g/dL without hemolysis, and partial response (PR) by Hb increase >2 g/dL or normalization of Hb with persistent hemolysis.8 Overall response (OR) was defined as the sum of PR and CR. Between 1995 and 2024, 344 patients with a biopsy-proven HHV-8 MCD were identified. Fifty-nine patients (17%) met AIHA criteria (AIHA+), with a median age at diagnosis of 45 years (Table 1). All patients had MCD flare at AHAI diagnosis, and all had detectable HHV-8 viremia. Thirty-eight patients (64%) were living with HIV (PLWH), including 10 patients who initiated antiretroviral therapy (ART) at the time of AIHA diagnosis. Eight AIHA+ patients (21% of PLWH) had undetectable HIV viremia. Median Hb values were similar between PLWH and HIV-negative patients (5.3 g/dL interquartile range; IQR 4.3–6.6 vs. 5.1 g/dL IQR 4.9–5.9, respectively), favoring a minimal direct role of HIV in the genesis of anemia. AIHA occurred during the first flare of MCD in most cases (n = 54/59, 92%), with a median time from AIHA onset to MCD diagnosis of 11 days (range: 8–65 days). No acquired TTP or dysglycemia was diagnosed. When compared to MCD patients without AIHA (AIHA−), AIHA+ patients were more often from Sub-Saharan ancestry (54% vs. 36%, P = 0.03). There were no differences between the two groups regarding HLH frequency, HIV infection, HHV-8 viremia, or KS at the time of AIHA diagnosis. At onset, the median Hb level was 5.2 g/dL (4.4–6.4), and the majority of AIHA+ patients presented with very severe (<6 g/dL) or severe (6.1–8 g/dL) anemia (respectively n = 33, 56% and n = 16, 27%, Table S1). RBC transfusions were required for 73% of patients, with a median of five packages transfused (3–6). DAT was strongly positive for IgG and C3d in 55 patients (93%), and for IgG only in 4 patients (7%). No cold agglutinin was found. Erythroblastic hyperplasia was found in 43% of bone marrow aspirations (10/23). Twenty-one AIHA+ patients (36%) fulfilled Hscore criteria for HLH,11 which may act as a confounding factor in the evaluation of AIHA parameters. However, in AIHA+ patients, the characteristics did not differ significantly between HLH− and HLH+ subgroups except for reticulocyte index (Table S2). Management and follow-up data were available for 49 AIHA+ patients (83%, Figure 1A). Nine patients (9/49,18%) received CS as monotherapy for at least 15 days (15–36 days) before MCD diagnosis or MCD suspicion, without any CR or PR, even transiently. After MCD diagnosis, 48 patients received specific treatment for an MCD flare (start day = day 0), which included etoposide (n = 42), liposomal doxorubicin (n = 4), or vinblastine (n = 2); 28 of these patients received concomitant corticosteroids (CS+ group), and 20 did not (CS− group). Virological status (including the frequency of HIV positivity, proportion of concomitant ART initiation, and HHV-8 viral load) as well as anemia parameters were quite similar in the CS+ and CS− groups (Table S3). This included the baseline Hb level, with median values of 5.2 g/dL versus 5.3 g/dL, respectively (Table S3). The frequency of overall response rate (ORR) at d30 did not differ significantly between subgroups (n = 24, 85% for CS+ vs. n = 19, 95% for CS−, P = 0.4), as well as the rates of CR and PR at d15 or d30 (Figure 1B). The median time to transfusion-free status was identical (6 days, Figure S1). The response patterns were similar in the HIV-positive and the HIV-negative group (OR 88% vs. 93%). Excluding patients meeting HLH criteria did not affect the results: the ORR at d30 was 100% in the CS− group and 89% in the CS+ group (P = 0.5). The majority of AIHA+ patients (n = 37, 76%) were given rituximab, with a median delay of 17 days (2–30) after they began treatment for their MCD flare. When focusing on the 19 patients who did not receive rituximab at d30, response rate remained comparable between CS− or CS+ groups (OR 73% in CS+ vs. 88% in CS−, P = 0.6). Among AIHA+ patients with extended follow-up (n = 49, median follow-up 74 months), 15 patients (31%) experienced AIHA relapse, occurring after a median delay of 23 months (10–35) following the initial episode. Four patients (8%) had multiple relapses. AIHA relapses were associated with MCD flares, except for one patient. The use of CS at AIHA diagnosis was associated with relapse (OR = 5.7, 95% CI 1.2–42, P = 0.04) in multivariate analysis (Table S4). Seven AIHA+ patients (14%) exhibited de novo or relapsed KS during follow-up (median time from AIHA diagnosis of 6 months). Three (6%) developed HHV-8+ lymphoma (median time: 25 months). Two-year OS was comparable between the AIHA+ (97.6%, 95% CI 93.1–100) and AIHA– groups (91.2%, 95% CI 87.7–94.9; log-rank P = 0.4, Figure S2). A total of ten AIHA+ patients deceased (20% vs. 22% in AIHA−, P = 0.9), with a median time to death of 66 months. The cause of death was known in eight cases, including bacterial infection (n = 3), MCD flare (n = 2), HHV-8+ lymphoma (n = 2), or KS (n = 1). While the incidence of KS was close in the patients who were treated with rituximab when compared to those who were not (n = 5/37, 14% vs. n = 2/11, 18%, P = 0.7), patients who received CS at AIHA diagnosis were more likely to develop KS during follow-up (6/28, 21%, vs. 1/20, 5% in the CS− group, P = 0.2). Our study presents the first dedicated analysis of AIHA specifically associated with HHV-8 MCD, an underreported complication of this lymphoproliferative disorder. We found that AIHA is a common manifestation in HHV-8 MCD, occurring as an initial complication in nearly 20% of newly diagnosed cases. AIHA exhibited a wAIHA pattern, with consistent detection of IgG autoantibodies. Most patients presented with severe anemia, a frequency nearly twice than reported in large primary wAIHA cohorts.12 While relapses occurred in approximately one-third of cases, AIHA manifestations were tied to MCD flares without evidence of chronic evolution. The outcome of AIHA in our series was favorable, with high response rates, including 90% of responders in the group that did not receive CS. This response rate is comparable to those reported in pilot studies evaluating rituximab for primary wAIHA where CS were coadministered.13-15 Importantly, none of the patients treated with CS alone before MCD treatment achieved remission in our study, suggesting that CS may not be the most appropriate monotherapy. Although the limited number of patients in subgroups prevented definitive conclusions, AIHA at MCD diagnosis does not seem to impact OS. However, KS incidence was four times higher in the follow-up of patients receiving CS. CS have been clinically linked to KS development in immunosuppressed individuals, and involved in a fourfold increase KS-related mortality in people living with HIV.16 Furthermore, CS use has been identified as an independent risk factor for hospitalization in patients receiving rituximab for wAIHA.17 These findings, together with the robust hematologic responses observed without CS in our cohort, collectively provide a rationale for steroid-sparing strategies in MCD-associated AIHA. Patients who experienced AIHA relapse were more likely to have received CS during the initial episode. This observation was not explained by baseline AIHA severity or rituximab use. One potential explanation is that CS might have impaired early T-cell responses against HHV-8-infected cells, contributing to higher relapse rates. The origin of autoantibodies associated with autoimmune complications of MCD remains elusive. During KSHV/HHV-8 MCD, infected plasmablasts (also known as viroblasts) found in blood or lymph node uniformly express monotypic IgM lambda immunoglobulin, despite having a polyclonal B-cell receptor (BCR). Autoimmune complications in MCD may directly arise from autoantibodies produced by viroblasts or, alternatively, through indirect bystander activation of B/plasma-cell compartments. The detection of IgG on RBCs supports the latter hypothesis, which is a mechanism also involved in TTP and autoimmune hypoglycemia associated with MCD.4, 5 Elevated levels of cytokines, particularly viral and human IL-6 and human IL-10, likely contribute to polyclonal B-cell stimulation and the emergence of autoreactive B- or plasma-cell clones.2 Despite potential unmeasured confounding factors, the CS+ and CS− subgroups were overall comparable. Nevertheless, our findings regarding CS therapy should be interpreted with caution. Beyond the retrospective design and small subgroup sizes—both emphasizing the need for larger prospective studies—the heterogeneity in rituximab use further limits interpretation. Importantly, stratification did not affect response rates. As AIHA responses to rituximab usually occur after 4–8 weeks,8, 18 and most patients received rituximab around 17 days after MCD flare treatment initiation, rituximab was unlikely to be the main driver of early responses. This supports a direct effect of first-line MCD therapy, primarily etoposide. Another potential bias that could have interfered with the interpretation of treatment response is the difference in diagnosis dates between the CS+ and CS− groups (March 2014 and March 2017, respectively), which may have led to heterogeneity in practices of antiretroviral treatment among HIV patients. However, the similar results obtained when restricting the analysis to non-HIV patients (95% in CS− and 85% in CS+ patients), together with the fact that treatment response criteria were assessed early (D15 and D30), make it unlikely that the initiation of ART alone accounted for the observed outcomes. The high prevalence of HLH in HHV-8 MCD complicates AIHA feature evaluation, but unchanged findings after excluding HLH cases suggest that HLH represents an independent complication. Similar to HLH, the identification of AIHA in MCD does not adversely affect survival when using a standard-of-care approach based on etoposide and rituximab.3 The higher proportion of Sub-Saharan ancestry also raises the possibility of co-existing hemoglobinopathies, that may also confound AIHA assessment, but systematic screening in most patients and the consistently strong DAT positivity support an autoimmune etiology. In conclusion, our findings establish AIHA as a frequent manifestation of HHV-8 MCD flares. Clinicians should be aware of this specific association and limit the use of steroids as first-line therapy. Yannick Dieudonné: Formal analysis; data curation; writing—original draft; writing—review and editing; investigation; conceptualization. Clément Gourguechon: Conceptualization; investigation; writing—review and editing. Romain Stammler: Writing—review and editing. Florence Delestre: Investigation. Boris Sorin: Investigation; writing—review and editing. Bertrand Dunogué: Investigation. Guillaume Dumas: Investigation. Guillemette Fouquet: Investigation. Anne-Marie Ronchetti: Investigation; data curation. Romain Paule: Investigation; data curation. Pascal Meliani: Investigation. Nathalie Pansu: Investigation. Rodolphe Buzele: Investigation. Robin Noel: Investigation; writing—review and editing. Jehane Fadlallah: Investigation. Marion Malphettes: Investigation. Emilie Corvilain: Investigation. Véronique Meignin: Investigation. Constance Guillaud: Investigation. Antoine Dossier: Investigation. Jérémie Dion: Investigation. Laurence Gérard: Investigation; writing—review and editing; methodology; data curation; formal analysis. Eric Oksenhendler: Investigation; writing—review and editing; validation; formal analysis; data curation. Lionel Galicier: Investigation; validation; writing—review and editing; formal analysis; data curation. David Boutboul: Conceptualization; investigation; writing—original draft; formal analysis; validation; supervision. The authors declare no conflicts of interest. All patients gave non-opposition to this study, and the research was approved by the local ethics committee and institutional review board. No specific funding was dedicated to this study. The data that support the findings of this study are available in the supplementary material of this article. 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.