Central nervous system involvement in blastic plasmacytoid dendritic cell neoplasm

TO THE EDITOR:

Blastic plasmacytoid dendritic cell neoplasm (BPDCN) is a rare and aggressive hematologic malignancy characterized by widespread cutaneous, blood, lymph node, and bone marrow involvement.¹ Before the modern targeted-therapy era, treatment has ranged from local therapy to multiagent chemotherapy regimens used in acute lymphoblastic leukemia (ALL), acute myeloid leukemia (AML), and lymphoma.^2,3 Several of these chemotherapeutic regimens have been successful in leading to a complete remission (CR) in patients, often followed by allogeneic or autologous hematopoietic stem cell transplantation (HSCT).^4,5 In BPDCN, historically, CR has been defined as an absence of blasts in the peripheral blood, <5% blasts in the bone marrow; hemoglobin concentration ≥9 g/dL without red blood cell transfusion for >2 weeks, platelet count ≥100 × 10⁹/L without platelet transfusion for >1 week, and absolute neutrophil count ≥1.5 × 10⁹/L.¹ Despite initial CR, overall survival (OS) is poor, and patients historically relapse within 2 years, commonly in skin, bone marrow, or central nervous system (CNS).^1,6 Recently, the emerging field of CD123-targeted agents in BPDCN have shown promising activity, including tagraxofusp-erzs (Food and Drug Administration [FDA] approved in December 2018 for patients with BPDCN aged ≥2 years),⁷ IMGN 632 (FDA Breakthrough Designation in October 2020 for relapsed/refractory BPDCN),⁸ and novel clinical trials with chimeric antigen receptor (CAR) T cells specifically dedicated to BPDCN including CD28/4-1BB CD123 targeting CAR T cells for use in patients with BPDCN,⁹ and the ongoing Mustang Bioscience CD123 CAR T-cell trial (www.clinicaltrails.gov, NCT04109482).¹⁰ 

The incidence of BPDCN CNS involvement, in particular of cerebrospinal fluid (CSF) positivity (+) from lumbar puncture (LP) analysis, is unknown in the modern targeted-therapy era. Notably, it has not been the standard of care to include LP at the time of diagnosis, and in most clinical trials, CNS disease has either been excluded, or prophylactic LPs have not been systematically performed.^11,12 Given the historically better outcomes of patients treated with regimens, including CNS prophylaxis^1,13,14 and HSCT,^4,5 BPDCN CNS involvement may occur more often than previously suspected. One study, performed before the use of targeted agents, demonstrated flow cytometric identification of CSF⁺ in 60% of patients with BPDCN at diagnosis and in 100% of patients at the time of relapse.¹⁵ Interestingly, many patients did not have neurologic symptoms, suggesting occult involvement and emphasizing the need for routine diagnostic screening for CSF⁺ disease. It is possible that CSF⁺ disease precedes systemic relapse and that targeted CNS prophylaxis would improve outcomes significantly.

We retrospectively analyzed the characteristics of patients with CSF⁺ BPDCN at our institution from 1999 through 2020 to better understand avenues for improved diagnosis and treatment. Among our 103 patients with BPCDN, 29 had undergone LP analysis, 13 (13%) had CSF⁺ diagnosed with frontline LP, and 23 (22%) had had CSF⁺ diagnosed during their BPDCN disease course. Among these 23 patients, 57% were routinely discovered with frontline LP, with the remaining 43% discovered in the relapsed setting after presenting with neurologic symptoms. The six remaining patients with negative LPs were asymptomatic. Importantly, because only 29 patients had ever undergone LP analysis, the CSF status of the remaining 74 patients with BPDCN in our cohort is unknown.

The median age of our CSF⁺ cohort was 63 years (range, 2-82), 74% were male, and 22% had prior/concomitant hematologic malignancies; Median time from BPDCN diagnosis to CSF⁺ discovery was 3.6 months (range, 0-36). Excluding those with upfront CSF⁺, median time from BPDCN diagnosis to CSF⁺ discovery was 5.8 months (range, 1-36; n = 18). Other sites of BPDCN involvement included skin (70%), lymph nodes (30%), and bone marrow (96%). Baseline laboratory values, cytogenetics, and molecular mutations of patients are included in Table 1. Significant differences included lower median baseline hemoglobin, higher frequency of TET2 mutations or variants, higher median baseline peripheral blood and bone marrow blast percentage in CSF⁺ vs CSF^- patients. Statistical analysis was performed with the Mann-Whitney U test for medians and with the χ² test for distributions.

Frontline therapies varied: HCVAD (hyper–cyclophosphamide, vincristine, doxorubicin, and dexamethasone; n = 8), CD123-targeting agent (n = 5), AML-based (n = 3), CHOP (cyclophosphamide, doxorubicin, vincristine, and prednisone+rituximab; n = 2), bortezomib-based (n = 1), hypomethylator-based (n = 1), venetoclax-based (n = 1), and other (n = 2). Thirty-four patients underwent allogeneic and 11 patients underwent autologous HSCT. Eleven patients with CSF⁺ disease (48%) underwent HSCT during treatment (10 allogeneic and 1 autologous), compared with 34 of 80 patients with CSF⁻ or unknown CNS involvement of disease (43%), who underwent HSCT. Three patients presented with isolated CNS relapse, all after induction therapy: (1 after modified HCVAD after HSCT, 1 after 3 + 7 therapy followed by high-dose cytarabine, and 1 after 3 + 7 therapy plus etoposide followed by AAL1131. HSCT improves median OS in both CSF⁺ (not reached vs 13 months; P = .005) and patients with CSF⁻ with BPDCN (40 vs 12 months; P = .013).

Twelve patients with CSF⁺ achieved CR1, and 50 patients with CSF⁻/unknown achieved C1. There was no significant difference in median OS between those with CSF⁺ and those with CSF⁻/unknown disease, nor was there a difference in median OS between those with CSF⁺ disease diagnosed in frontline tests and all others, including those with later CSF⁺ relapse (Figure 1). There was a nonsignificant trend toward improved survival in those with frontline-diagnosed CSF⁺ disease. This finding could generate the hypothesis that upfront intrathecal therapy would benefit patients with known CSF⁺ disease, but the small sample size and insignificant P-value do not allow us to draw any definitive conclusions.

We have observed the emerging phenomenon of CSF⁺ on more LPs than expected in patients with BPDCN, akin to 5% to 17% found in high-risk ALL^16-18 or 13% to 17% found in Burkitt’s leukemia/lymphoma.¹⁹ These rates are higher than would be expected for patients with AML or other myeloid malignancies.^20,21 Martín-Martín et al demonstrated high rates of CSF⁺ in BPDCN, mostly in asymptomatic cases.¹⁵ Now, with the advent of targeted therapies in both frontline and relapsed settings, the natural history of BPDCN is being altered favorably, with patients living much longer, perhaps long enough to have CNS relapses. This phenomenon is reminiscent of the late CNS relapses observed in Philadelphia⁺ B-ALL, where increased frequency of prophylactic LPs positively influences late-term outcomes.^22,23 The high rate of CSF⁺ recurrence among patients with BPDCN may explain the high rate of relapse despite initial CR with treatment, if CNS prophylaxis was not used. Despite these findings, to date, it has not been standard to include systematic LP screening or prophylactic intrathecal chemotherapy for all patients with BPDCN.¹¹ 

Of our patients, 13% were diagnosed with CSF⁺ BPDCN by frontline routine LP, and 57% who had ever had CSF⁺ disease were diagnosed by frontline LP. Importantly, only 29 patients of our entire cohort of 103 patients actually underwent LP; therefore, this rate of asymptomatic CSF⁺ in the frontline setting may actually have been higher if all patients had received LPs regularly. Additionally, 96% of patients with CSF⁺ disease also had bone marrow involvement of BPDCN, with a median bone marrow blast percentage significantly higher than in those with CSF⁻ disease. BPDCN with bone marrow involvement often behaves clinically like high-risk acute leukemia,¹ suggesting that those with higher disease burden are more likely to have CNS involvement than are those with skin-limited disease. Patients with CSF⁺ disease had more complex cytogenetics, and all demonstrated the presence of TET2 mutation/variant, compared with those with CSF⁻ disease or unknown BPDCN CNS involvement. TET2 is one of the most commonly mutated genes in various hematologic malignancies,²⁴ and the prognostic implications of TET2 in patients with BPDCN warrants further study.

Our findings demonstrate a need for earlier and more frequent performance of diagnostic and therapeutic LPs in patients with BPDCN. At our institution, we have therefore designed the Triple Therapy Program consisting of tagraxofusp, venetoclax, and HCVAD/mini-CVD (age-adjusted) to include routine screening LPs (www.clinicaltrials.gov, NCT04216524). This triple-therapy regimen originally consisted of intrathecal chemotherapy in cycles 3, 5, and 7, with alternating cytarabine and methotrexate.²⁵ Moving forward, we will incorporate earlier LPs with prophylactic intrathecal chemotherapy during cycles 1 and 2. With these new recommendations, clinicians treating BPDCN must continue to closely monitor for neurotoxicity secondary to intrathecal chemotherapy.²⁶ Future areas for study include evaluation of the newer CD123-targeted agents,^7,8 BCL-2–targeting agents,^12,27 and CAR-T cells,⁹ with regard to crossing the blood-brain barrier and addressing the need for prophylactic screening LPs during therapy in all patients with BPDCN. These studies will help guide future recommendations for timing of diagnostic and therapeutic LPs in evaluating patients with BPDCN for CNS disease involvement, and the importance of incorporating this paradigm into each of the future BPDCN-directed systemic therapies and novel clinical trial approaches.