Cyclophosphamide as a first-line therapy in LGL leukemia.

Large granular lymphocyte (LGL) leukemia is a T or NK clonal disorder characterized by the tissue invasion of marrow, spleen and liver. Clinical presentation is dominated by recurrent infections associated with neutropenia, anemia, splenomegaly and autoimmune diseases, particularly rheumatoid arthritis.[1, 2, 3, 4] Recently, STAT3 and STAT5 mutations have been detected in T-LGL and in NK-LGL leukemias.[5, 6, 7, 8] These somatic mutations, coupled to other intrinsic and extrinsic mechanisms, are likely to induce constitutive activation of the JAK/STAT pathway thus contributing to maintenance of leukemic LGL survival.[9] These findings strongly suggest a common specific pathogenic pattern in T-LGL and NK-LGL leukemias and provide justification for consideration of the same treatment options. Indications for treatment are severe or symptomatic neutropenia, symptomatic or transfusion-dependent anemia or associated autoimmune diseases requiring therapy. There is no standard treatment for patients with LGL leukemia. All the largest series published in the literature (collecting data on more than 40 patients) are retrospective. Data are very heterogeneous and treatment outcome per single agent is available for very few patients. Immunosuppressive therapy remains the foundation of treatment including single agents that is, methotrexate, oral cyclophosphamide or cyclosporine. On the basis of an initial study showing very good overall response rate (ORR) using methotrexate, this drug has remained the most recommended option in LGL leukemia.[10] Oral low dose cyclophosphamide was first used in pure red cell aplasia associated with LGL leukemia.[11, 12] In a French series, cyclophosphamide was shown to be also efficient in neutropenic patients and for those who failed methotrexate.[13] Those results suggested that cyclophosphamide used as first-line therapy could be an interesting alternative to methotrexate.

In this letter, we describe the encouraging results of cyclophosphamide used in a series of 45 previously untreated LGL leukemia patients. Patients suffering from LGL leukemia and treated with cyclophosphamide as first-line therapy were included in this retrospective study. Patients were screened from the Italian, French and USA Penn State registries. Patients gave their informed consent for data collection. The diagnosis of LGL leukemia was based on a chronic LGL peripheral blood expansion (>0.5 × 109/l), usually lasting for more than 6 months. Criteria for T-LGL leukemia included expression of LGL surface markers compatible with a typical T-cell (commonly αβ+ or γδ+/CD3+/CD8+/CD57+ and/or CD16+) phenotype associated with clonal rearrangement of TCRγ gene using PCR or clonal Vβ expression using flow cytometry. Criteria for NK-LGL lymphocytosis/chronic NK-LGL leukemia included expression of LGL surface markers compatible with a NK-cell (commonly CD3−/CD8+/CD16+ and/or CD56+) phenotype with more than 0.75 × 109/l circulating cells.[14, 15] Response to treatment was determined periodically on blood cell count and only best response was taken into account. Hematological complete response (CR) was defined by a normal blood count (hemoglobin (Hb)>12g/dl, platelets>150 × 109/l, absolute neutrophil count (ANC)>1.5 × 109/l and lymphocytosis<4 × 109/l) and LGL peripheral count in a normal range (<0.3 × 109/l). Molecular CR was based on hematological CR associated with a negative PCR analysis for CD3+ cases. Hematologic partial response was defined as an improvement in blood count specified as follows: ANC increasing more than 50% and reaching more than 0.5 but less than 1.5 × 109/l; Hb level increasing more than 2 g/dl and transfusion independent without reaching 12 g/dl level. Treatment failure was defined as a progressive disease (worsening of cytopenia or organomegaly) or a stable disease (none of the later given criteria met). Some patients received cyclophosphamide because of symptoms not related to cytopenia. For those patients, response criteria included clinical symptom resolution. Patients who received prednisone, granulocyte colony-stimulating factor or erythropoiesis-stimulating agent before or at the same time of cyclophosphamide were included in this retrospective study. For the descriptive analysis, qualitative variables were described using numbers and percentage, whereas medians and extremes were used to describe quantitative analyses. Qualitative variables were compared using χ2 or Fisher's test.

A total of 45 patients treated with cyclophosphamide as a first-line therapy for LGL leukemia from 1989 to 2012 were retrospectively included in this series. Clinical characteristics are described in Table 1. Starting doses were as follows: 100 mg/day (n=36), 50 to 75 mg/day (n=8) taken orally and 1 g IV/month (n=1). Median time from diagnosis to treatment was 3 months (range, 0–55). Treatment was initiated because of severe isolated neutropenia (n=16, 36%), neutropenia and anemia (n=2, 4%), transfusion-dependent anemia (n=15, 33%) and thrombocytopenia (n=5, 11% including three cases of idiopathic thrombopenic purpura). Seven non-cytopenic patients (15%) were treated for disease-associated LGL leukemia: neuropathy (n=3), vasculitis (n=1) and constitutive symptoms (n=3). ORR was 71% (32/45): there were 21 CR (47%) including three molecular responses (mCR) and 11 (24%) partial responses. ORR was 72% versus 68% for T-LGL and NK-LGL subtype, and 72% versus 67% for neutropenic and anemic patients, respectively (P=not significant). Patients treated for symptoms related to LGL leukemia had a 94% ORR (6/7). Eighteen patients (40%) were initially co-treated with prednisone and no significant impact was found for the response to cyclophosphamide (P=0.31). As well as, concomitant granulocyte colony-stimulating factor administration with cyclophosphamide did not influence the time to response and the ORR. The median time to reach best response was 4 months (range 0.8–21) and patients were treated for a median time of 6.4 months (range 0.5–33). Evidence of clinico-biological improvement was systematically observed within the 4 months following treatment initiation. Therefore, we assume that cyclophosphamide should be given for at least 4 months before changing drug regimen. Median treatment duration was 8.5 months (range 3.4–33) for responders, except one patient who received several courses of cyclophosphamide for 7 years. With a median time follow-up of 35 months (range 3.8–277), four out of the 32 responders relapsed (13%): one of these patients responded again to a short course of cyclophosphamide and was maintained subsequently on cyclosporine decided thereafter. The three others were switched to either methotrexate or cyclosporine. Eight patients experienced grade 1–2 toxicities (18%). Only three patients (7%) stopped treatment: one because of worsened anemia and two because of febrile neutropenia not obviously related to hematological toxicity. One patient experienced temporary worsening of anemia without stopping therapy.

Table 1

Patients characteristics, n=45

Median age	60 (24–90)
Sex ratio (M/F)	24/21
T-LGL	32
NK-LGL	13
Blood counts
Median lymphocytes (109/l)	5.1 (0.5–26.8)
Median LGL (n=34)a (109/l)	3.34 (0.5–25.2)
LGL>4 × 109/l	21/34 (62%)
ANC<0.5 × 109/l	11/45 (24%)
Hb<11 g/dl	24/45 (53%)
Hb<8 g/dl	7/45 (16%)
Platelets<100 × 109/l	10/45 (22%)
LGL marrow infiltration (n=40)b	29/40 (73%)
Autoimmune cytopeniasc	n=11
PRCA	4d
AIHA	4
ITP	3
STAT3 mutation	7/20 (35%)
Autoimmune disorders	n=11
Rheumatoid arthritis	2
Hypothyroidism	4
Others	5

Abbreviations: ANC, absolute neutrophil count; F, female; Hb, hemoglobin; LGL, large granular lymphocyte; M, male.

Median LGL count was available for 34 patients.

Bone marrow infiltration status was only available for 40 patients, assessed on bone marrow aspirate or bone marrow biopsy using the Morice WG criteria (diffuse interstitial pattern with small clusters of CD8+, TIA-1+ and granzyme B+ cells).[16]

PRCA, pure red cell aplasia; AIHA, autoimmune hemolytic anemia; ITP, idiopathic thrombopenic purpura.

One patient had also Crest syndrome.

We previously mentioned that cyclophosphamide induces a very good ORR in LGL leukemia in de novo or relapsing patients.[14] The ORR of 71% described in our series confirmed and emphasized what has been reported in the literature with a total of 25 responders out of 38 patients treated as first line with cyclophosphamide (66% ORR) (Table 2).[11, 12, 13, 17, 18] We show that cyclophosphamide compares favorably to methotrexate given as a first-line therapy. In 1993, Loughran et al.[10] reported a 60% ORR in a prospective series of 10 patients receiving methotrexate at a weekly dose of 10 mg/m2. These results were less encouraging in two larger series. One comes from the prospective ECOG study showing a 37% ORR in 56 patients and the other is retrospective from the French registry and reported a 44% ORR in 36 patients.[13, 19] Furthermore, molecular response is rarely obtained and the incidence of relapse following methotrexate is, at least in French experience, estimated at 67%.[13] Our series demonstrates that cyclophosphamide used as first-line therapy is effective in T/NK-LGL leukemia, in both neutropenic and anemic patients. ORR and response duration seem encouraging. We recommend only 9–12 months of treatment. It seems sufficient to induce durable remissions and to avoid the complication of myelodysplastic syndromes/acute myeloid leukemia, which although rare is dependent on cumulative dose and length of exposure. For responding patients, tapering dose to 50 mg per day would be a reasonable option. Although recognizing the limits of a retrospective study, we suggest that cyclophosphamide could be an interesting alternative to methotrexate as first-line therapy in LGL leukemia. A prospective study comparing cyclophosphamide to methotrexate as first-line therapy is currently ongoing in France.

Table 2

Summary of results for cyclophosphamide used as a first-line therapy in patients with LGL leukemia

Number of patients	ORR	Complete remission	Reference
16	10 (63%)	6	Dhodapkar[17]
5	4/5 (80%)	4	Go[11]
8	6/8 (75%)	2	Fujishima[12]
4	3/4 (75%)	2	Bareau[13]
5	2/5 (40%)	Unknown	Mohan[18]
45	32/45 (71%)	21	Our series