Phase 2 trial of rituximab with either pentostatin or bendamustine for multiply relapsed or refractory hairy cell leukemia Open Access

• Durable CRs and MRD-free CRs were achieved with pentostatin-rituximab (DCFR) and bendamustine-rituximab (BR).

• Possible superiority of DCFR was hypothesis-generating and might mainly apply to patients less pre-treated and with lower disease burden

The primary objective in multiply-relapsed hairy cell leukemia and variant (HCL/HCLv) was to determine if pentostatin-rituximab (DCFR) and bendamustine-rituximab (BR) each have overall response rate (ORR) exceeding that historically achieved by rituximab alone (~40%) in favor of 65%. Prospective data was unreported for either regimen. Fifty-six patients received six 28-day cycles of rituximab (375 mg/m² days 1, 15) with either bendamustine (90 mg/m² days 1, 2) or pentostatin (4 mg/m² days 1, 15). Eligibility required ≥2 purine analogs, or one plus rituximab for initial response <1 year. Complete remission (CR) and minimal residual disease (MRD) were studied, and progression-free survival (PFS) with hazard ratios (HR) determined. Although patients were assigned to either regimen through randomization for increasing homogeneity of the two treatment groups, the DCFR arm had fewer prior purine analogs (p=0.021) and lower baseline marrow HCL/HCLv infiltration (p=0.013). ORRs for DCFR and BR were 93% and 86%, 95% confidence intervals (95%CI) 83-102% and 73-99%, each exceeding the primary objective (40% in favor of 65%, p<0.0001 for each). Rates for CR and MRD-free CR and median PFS (141 vs 50 months, HR 0.63, 95%CI 0.32-1.25) numerically favored DCFR but that arm was significantly enriched with lower prior purine analogs and marrow infiltration each of which was associated post hoc with better response. Post hoc subgroup analysis particularly for the 41 patients with classic HCL suggested any superiority of DCFR vs BR might apply to the more favorable patients. DCFR and BR were highly effective in multiply relapsed HCL/HCLv. Possible DCFR superiority was hypothesis-generating, given uneven baseline risks and trial design. NCT01059786.