Higher rate of progression in HIV^– than in HIV⁺ patients after rituximab for HHV8⁺ multicentric Castleman disease

Kaposi sarcoma (KS)–associated herpesvirus/human herpesvirus 8 (KSHV/HHV8)–associated multicentric Castleman disease (HHV8⁺ MCD) is a rare lymphoproliferative disorder associating lymphadenopathy, splenomegaly, fever, and inflammatory symptoms with high serum C-reactive protein (CRP), hypergammaglobulinemia, and a high blood HHV8 DNA viral load.^1-3 HHV8⁺ MCD mainly affects patients who are HIV infected (HIV⁺) but can also be observed in those who are not HIV infected (HIV^–), such as patients born in countries with high HHV8 infection prevalence, men who have sex with men, and organ transplant recipients.^4,5 

In the setting of HIV infection, rituximab-based therapy has dramatically improved the 5-year overall survival (OS) >90%.^6,7 KSHV/HHV8-infected viroblasts are polyclonal naive B cells expressing a restricted surface immunoglobulin Mλ phenotype. They exhibit a plasmacytic differentiation profile distinct from that of classic plasmablasts but with a low level of CD20 expression.^8-11 However, in 2 open-label studies, rituximab has been shown to induce complete remission in most patients.^12,13 Rituximab can be used as a single agent weekly for 4 weeks or associated with etoposide in patients at high risk with hemophagocytic syndrome or organ failure or to avoid a possible flare of the disease during rituximab therapy.⁵ However, with the use of rituximab, worsening KS may occur in patients with baseline KS lesions. The combination of rituximab with liposomal doxorubicin appears to avoid or attenuate KS exacerbation.¹⁴ In addition, rituximab-based therapies have also reduced the risk of HHV8-associated lymphoma.¹⁵ In a previous study in patients who were HIV⁺, MCD relapse occurred in 23% of the patients, with a median time to the first relapse of 30 months. All patients could be successfully retreated at relapse with rituximab-based therapy.¹⁶ We have previously shown that rituximab with or without etoposide appears to be as effective in the population who are HIV^– than in those who are HIV⁺.⁵ 

Here, we first looked at factors that could predict disease progression and compared the progression rates in patients who were HIV⁺ and HIV^–. In addition, we studied the kinetics of 3 biomarkers that could predict relapse: blood HHV8 DNA, which has already been shown to predict relapse¹⁷; B-cell restoration reflecting the exhaustion of the rituximab-induced B-cell depletion; and serum CRP level, a major marker of HHV8⁺ MCD attack.

Patients were enrolled in the registry from the National Reference Center for Castleman disease. All patients gave informed consent, and the registry was approved by the local ethics committee and institutional review board. Enrolled patients had been diagnosed with biopsy-proven HHV8-associated MCD or KSHV inflammatory cytokine syndrome (KICS), tested for HIV between 1 January 2000 and 31 December 2020, and treated with rituximab-based therapy as a first- or second-line therapy. For this study, we then selected the patients who achieved complete remission after rituximab therapy, excluding patients with initial treatment failure. A complete response to rituximab was defined as the resolution of constitutional symptoms, including fever, regression of peripheral lymphadenopathy and splenomegaly, correction of cytopenia (hemoglobin > 10g/dL and platelets > 100 × 10⁹/L), and serum CRP level below 20 mg/L, and the absence of requirement of additional therapy. Computed tomography or positron emission tomography–computed tomography were not systematically performed in patients who were asymptomatic. Clinical and biological data, including CD19⁺ B-cell count using flow cytometry, whole blood HHV8 DNA using polymerase chain reaction, and serum CRP, were collected during follow-up.

Disease progression was defined as the occurrence of MCD relapse (recurrence of similar symptoms associated with a high HHV8 viral load), KS progression, lymphoma requiring specific therapy, or death. MCD relapse was defined per previously established criteria for an attack, even when not histologically proven.¹³ 

The patient’s characteristics were analyzed using descriptive statistics. Categorical variables were expressed as numbers (percentages) and continuous variables as medians with an interquartile range (IQR). Comparison of variables between groups for nominal variables was done using χ² or Fisher exact test, and for continuous variables, comparison was done using nonparametric Mann-Whitney test. Progression-free survival (PFS) was defined as the time from rituximab therapy until failure or censoring the data of patients still alive at the final visit. The OS was analyzed via the Kaplan-Meier method, and the nonparametric log-rank test was used to compare univariate survival distributions. Baseline covariates at diagnosis were tested using the Cox proportional hazards model to determine their effect on PFS. Hazard ratios are reported with a 95% confidence interval (CI). For the multivariate analysis, the regression model was set with those covariates that yielded P < .10, using univariate analysis. A second multivariate model included time-varying covariates. Statistical analysis was performed using STATA/IC 16.1 (College Station, TX), and the significance level was set to 5%, with no adjustments for multiplicity.

Between 1 January 2000 and 31 December 2020, 113 patients with HHV8-associated MCD were treated with a rituximab-based regimen as a first or second line of treatment and were eligible for inclusion. Among them, 14 were excluded because of missing data at baseline, and 99 patients (73 who were HIV⁺ and 26 who were HIV^–; median age, 45 years) were included in the study, with a strong male predominance in both groups. Some of these patients have previously been reported in series from our center.^4,5 Three patients did not have a lymph node biopsy specimen but presented with KICS¹⁸ and were included in the study. Patient evaluation at HHV8⁺ MCD diagnosis revealed similar characteristics in the populations that were HIV⁺ and HIV^–, although patients who were HIV^– were older, had normal CD4⁺ T-cell counts, and had less frequent KS (Table 1). The median serum CRP level was 120 mg/L (IQR, 49-200), and the median blood HHV8 DNA level was 5.11 log copies per mL (IQR, 4.2-5.8), with similar values in HIV⁺ and HIV^– groups. For the purpose of the study, we focused on the 95 patients (96%) who achieved remission after rituximab-based therapy (supplemental Figure 1).

Forty-one patients (43%) had received rituximab (off-label) as a first-line therapy, whereas 54 patients had received etoposide before rituximab therapy. At baseline (rituximab therapy), 66 patients (69%) still had active MCD. Single-agent rituximab therapy (375 mg/m² IV every week for 4 weeks) was administered to 17 patients (18%), whereas combination therapy was administered to 78 patients (82%). The most frequent combination (n = 68) associated rituximab and etoposide (100 mg/m² etoposide after each rituximab infusion; Table 2).

After a median follow-up of 51 months (IQR, 24.5-90), with a median number of 6 visits per patient (IQR 4-8), persistent remission was observed in 59 patients (62%), whereas progression was observed in 24 of the 70 patients who were HIV⁺ (34%) and 12 of the 25 who were HIV^– (48%). Most of the 36 disease progressions were related to MCD progression (n = 25). Four patients experienced active KS lesions requiring specific therapy, 3 patients developed HHV8-associated lymphoma, and an additional 4 patients had died (1 from leukemia and 3 from an unknown cause). The median time to progression was 101 months. The 2-year and 5-year PFS were 75.8% (95% CI, 65.0-83.7) and 54.4% (95% CI, 40.8-66.0), respectively (Figure 1A). At MCD relapse, most patients exhibited similar, although slightly less severe, features compared with the features observed at diagnosis (supplemental Table 1). Among the 25 patients who had experienced a first relapse, 19 received a second cycle of 4 rituximab infusions, and all (100%) achieved remission. Twelve patients (55%) presented with multiple relapses (n ≥2).

The median time for the first CD19⁺ B-cell count >1 × 10⁶/L was 13.6 months, and the median 2-year probability for such an event was 55.8% (95% CI, 45.8-66.3), with no difference between the HIV^– and HIV⁺ groups: 59.4% (95% CI, 40.1-79.6) and 54.5% (95% CI, 43.1-66.7), respectively.

Blood HHV8 DNA levels rose above 3 logs copies per mL in 57 patients (60%). The median time for the first HHV8 DNA >3 log copies per mL was 20 months, and the 2-year probability for such an event was 54.2% (95% CI, 44.0-65.0), with no significant difference between the HIV^– and HIV⁺ groups: 60.5% (95% CI, 40.8-89.3) and 51.8% (95% CI, 40.3-64.2), respectively.

Forty-two patients (44%) experienced a rise in the serum CRP level above 20 mg/L during follow-up. The median time for CRP >20 mg/L was 59 months, and the 2-year probability for such an event was 34.5% (95% CI, 25.4-55.6), with no significant difference between the HIV^– and HIV⁺ groups: 43.1% (95% CI, 25.1-66.7) and 31.6% (95% CI, 21.8-44.3), respectively.

We evaluated the impact of baseline covariates on the risk of progression after rituximab-induced complete remission. In a univariate Cox model, only 2 covariates were associated with a higher risk of progression: hemoglobin levels below 10 g/dL and HIV-negative status. These 2 covariates remained independently associated with a higher risk of progression in a multivariate model (P = .01 and P = .04; respectively). The risk of progression was higher in the HIV^– group than in the HIV⁺ group, with a 5-year PFS of 25.7% (95% CI, 5.2-53.6) and 61.6% (95% CI, 46.2-73.7), respectively (Table 3; Figure 1). Of note, when the analysis was restricted to the population that was HIV⁺, none of the HIV–associated markers (plasma HIV RNA and CD4⁺ cell count) had a significant impact on the risk of progression (data not shown).

We then incorporated 3 time-dependent covariates in the model (B-cell count ≥ 1 × 10⁶/L, blood HHV8 DNA ≥ 3 log copies per mL, and serum CRP ≥ 20 mg/L). In the multivariate analysis, HIV-negative status, HHV8 DNA ≥3 log copies permL and CRP ≥ 20 mg/L remained independently associated with a higher risk of progression (P = .001; P = .01; and P = .01; respectively; Table 3). The median time for progression after the first detection of blood HHV-8 DNA > 3 log copies per mL was 25.9 months, with a 2-year probability of 48.1% (95% CI, 35.4-62.5). The 2-year progression rate was 40.9% (95% CI, 27.0-58.6) in the HIV⁺ group and 66.0% (95% CI, 41.6-88.4) in the HIV^– group. At last follow-up, the median HHV8 viral load for the whole population was below 2 log copies per mL, below 2 logs per mL for the 59 patients who had not experienced progression, and 4.81 logs per mL for the 36 patients who had progressed during follow-up. The median time for progression after the first detection of serum CRP >20 mg/L was 4.8 months, with a 2-year probability of 63.7% (95% CI, 48.7-78.5; supplemental Figure 3).

There was a trend for a higher rate of progression in patients who were treated with rituximab alone (10 of 17 [58.8%]) than in patients who received rituximab in association with chemotherapy (26 of 78 [33.3%]; P = 0.45).

Ten patients had died, including 4 patients who died after their data were censored for a previous event. The median OS was not reached. The estimated 5-year OS was 84.4% (95% CI, 69.7-92.3), with no significant difference between the populations that were HIV^– and HIV⁺ (P = .17; supplemental Figure 4).

This study provides new insights on the outcome of HHV8⁺ Castleman disease in a large population of patients who were HIV⁺ or HIV^– and treated with rituximab. To our knowledge, this is the first series evaluating rituximab therapy in patients who were HIV^–. Rituximab-based therapies have dramatically improved the prognosis of HHV8⁺ MCD and reduced the risk of developing HHV8-associated lymphoma in the context of HIV infection. However, despite this huge therapeutic advance and the excellent OS achieved (above 90% at 5 years), HHV8⁺ MCD remains a relapsing-remitting disease. The present report describes the patterns of progression after rituximab-based therapy not only in patients who were HIV⁺ but also in those who were HIV^–. The population that was HIV⁺ is very similar to that a large cohort series in the United Kingdom and United States in terms of demographics and immune deficiency at baseline. However, in the series from the United States, almost all patients had an undetectable HIV viral load at baseline, and most patients were treated with rituximab in association with liposomal doxorubicin. In many patients from the current series, the diagnosis of MCD was simultaneous with the diagnosis of HIV infection, explaining why only 25% had already achieved complete HIV replication control under combination antiretroviral therapy (cART).

The overall proportion of patients who experienced progression (34.2%), whatever its type, MCD relapse (21.4%), KS progression (5.7%), occurrence of lymphoma (1.4%), or death (5.7%), was slightly higher than that previously described by Pria et al¹⁶ and similar to that observed by Ramaswami et al.¹⁹ When restricted to MCD relapse (n = 25), the 5-year relapse-free survival was 66.7% (95% CI, 53.5-76.9). The clinical presentation of HHV8⁺ MCD at relapse was similar to that at the first episode of MCD with milder symptoms, presumably resulting from greater vigilance by both patients and physicians.

At diagnosis, patient evaluation revealed similar characteristics in the populations that were HIV⁺ and HIV^–, although patients who were HIV^– were older, had normal CD4⁺ T-cell counts, and had less frequent KS. Even with a shorter follow-up period, the probability of progression was significantly higher in the population that was HIV^–. The older age of the patients who were HIV^– is probably not an explanation for this poorer outcome, as age did not appear to be associated with prognosis in the univariate analysis. In contrast, the role of immunological improvement after the introduction of cART in most of these patients is a plausible explanation. Only 23% of the patients who were HIV⁺ were receiving cART at MCD diagnosis, and most had uncontrolled HIV infection and immune deficiency with a median CD4 cell count of 233 × 10⁶/L. All of these patients had started receiving cART together with rituximab therapy for MCD. Interestingly, patients with an HIV viral load <50 copies per mL at baseline did worse than patients who had started receiving cART simultaneously with rituximab therapy. Nine out of 17 patients (53%) experienced disease progression, whereas only 15 out of 53 patients (28%) with initially uncontrolled HIV infection had disease progression during follow-up after being treated with rituximab and cART. Therefore, HIV replication control and subsequent immune restoration might have contributed to the control of HHV8 infection and a lower risk of progression while receiving cART. In contrast, the possible underlying immune defect associated with the occurrence of HHV8⁺ MCD in patients who were HIV^– may worsen over time in this older population.

To better understand the patterns of progression after rituximab-based therapy, we also evaluated 3 biomarkers that could be involved during the post-rituximab therapy follow-up: (1) restoration of circulating B cells subsequent to exhaustion of the rituximab-induced B-cell depletion; (2) recurrence of high levels of blood HHV8 DNA that has been shown to predict relapse¹⁷; and (3) rise in serum CRP level that could be the first symptom of a MCD attack. The 2-year probability of observing the recirculation of B cells was >50%. However, the B-cell restoration was not predictive of disease progression. In contrast, the 2-year probability of a rise in blood HHV8 DNA above 3 logs per mL was also >50% but was strongly associated with a risk of progression with a 2-year PFS of 48%. The HHV8 viral load in patients with asymptomatic HHV8 infection is usually <3 logs, whereas it is ∼3 logs for patients with KS and ∼5 logs for patients with MCD. Therefore, an arbitrarily chosen cut off of 3 logs is reasonable. The rise of serum CRP above 20 mg/L was also highly predictive of treatment failure, because the 1-year probability for progression after such an event was 54%, suggesting that both blood HHV8 viral load and serum CRP represent interesting markers to predict progression and could, therefore, be monitored during follow-up.

The limitations of our study are related to the retrospective design and the absence of planned timing for blood samplings during follow-up. However, most patients were followed up in a single center by physicians trained in HHV8⁺ MCD. This might have minimized heterogeneity in the monitoring during follow-up. There are also several strengths to this study. This is the first series to include patients who were HIV^– with MCD and treated with rituximab, highlighting the high risk of progression in this population. In addition, the kinetics of HHV8 viral load and serum CRP provide interesting information regarding the pattern of progression in these patients and help physicians in the management of this rare, although potentially lethal, condition.

The authors thank Constance Delaugerre for assistance in the virological tests and Mirlinda Berisha for data management.

Contribution: V.M. reviewed the pathological specimen; C.R. and E.O. wrote the manuscript; J.F., E.C., L. Galicier, and D.B. took care of the patients; and L. Gérard performed the statistical analysis.

Conflict-of-interest disclosure: E.O. serves as a consultant for EusaPharma and CSL Behring. The remaining authors declare no competing financial interests.

Correspondence: Eric Oksenhendler, Département d’Immunologie Clinique, Hôpital Saint-Louis, 1 Ave Claude Vellefaux, 75010 Paris, France; e-mail: eric.oksenhendler@aphp.fr.