Background:
* Hairy cell leukemia (HCL) is highly responsive to purine analogs cladribine and pentostatin, without evidence of cure. Neither is standard after 2 courses, due to cumulative marrow and T-cell toxicity and declining remission rates and durations. Once resistant, patients after multiple relapses can die of disease-related cytopenias.
* Rituximab alone in 51 patients from 5 trials who had cytopenias and at least 1 prior purine analog resulted in 10 complete + 10 partial remissions (complete response (CR) + partial response (PR) = overall response rate (ORR 39%).
* Rituximab with cladribine gives high CR rates in 1st or 2nd line but is not standard.
* While cladribine use is more common for 1st and 2nd line, pentostatin is often used for subsequent treatment because of \< 100% cross-resistance.
* Retrospective published data for pentostatin plus rituximab in HCL include 7 of 7 responses with 6 (86%) CRs, and there are no prospective data.
* Recombinant immunotoxins targeting cluster of Differentiation 25 (CD25) (LMB-2) and cluster of differentiation-22 (CD22) (CAT-3888 (BL22) and R490A (HA22) are highly active in purine analog resistant HCL. Palliative pentostatin-rituximab is often used off-protocol for patients with immunogenicity needing more therapy.
* Bendamustine is approved for early treatment of chronic lymphocytic leukemia (CLL) and is effective with rituximab for relapsed/refractory CLL. Its use in HCL is unreported.
* CRs with minimal residual disease (MRD) by immunohistochemistry (IHC) of bone marrow biopsy (BMBx IHC), can relapse early. Tests for HCL MRD in blood or marrow include flow cytometry (fluorescence-activated cell sorting (FACS) or polymerase chain reaction (PCR) using consensus primers. The most sensitive MRD test in HCL is real-time quantitative PCR using sequence-specific primers (real-time quantitative polymerase chain reaction (RQ-PCR).
* Of 5 HCL-specific trials listed on Cancer.gov, 2 are phase II trials of cladribine + rituximab in 1st and 2nd line (1 randomized at National Institutes of Health (NIH), 1 non-randomized at MDA), and 3 NIH phase I-II trials of recombinant immunotoxins BL22, HA22 and LMB-2.
Objectives:
-Primary:
--To determine if pentostatin + rituximab and bendamustine + rituximab are each associated with adequate response rates (ORR=PR+CR) in patients with relapsed HCL, and, if so, to select which combination is likely to be superior.
Eligibility:
* HCL needing therapy, either greater than or equal to 2 prior courses of purine analog, 1 course purine analog plus greater than or equal to 1 course rituximab if \< 1 year response to the 1 course purine analog, diagnosis of HCL variant (HCLv), or unmutated immunoglobulin heavy variable 4-34 (IGHV4-34+) expressing HCL/HCLv.
* Prior treatment, ineligibility for, or patient refusal of recombinant immunotoxin.
Design:
* Rituximab 375 mg/m\^2 on day 1, 15 for 6 x 28-day cycles (all 72 patients).
* Initial tolerability study: 12 patients receive rituximab + bendamustine (nonrandom), including 6 at 70 mg/m\^2 and 6 at 90 mg/m\^2 of bendamustine.
* Randomize: 1) 28 patients to bendamustine 90 mg/m\^2/day, days 1 and 2 each cycle 2) 28 patients to pentostatin 4 mg/m\^2 days 1 and 15 of each cycle.
* Non-randomized: up to 4 patients to receive either bendamustine 90 mg/m\^2P2P/day, days 1 and 2 each cycle or pentostatin 4 mg/m\^2P2P days 1 and 15 of each cycle.
* Statistics: If \> 14/28 respond, can conclude with 90% power that response \> 40% in that arm. \>80% probability of selecting the better arm if true response probability is approximately 40-50% on the inferior arm and \>15% higher on the superior arm.
* Stratify to equalize the % of patients/arm refractory to last course of purine analog.
* Accrual Ceiling: 72 evaluable participants
