Phase I Study of the Investigational Aurora A Kinase Inhibitor Alisertib plus Rituximab or Rituximab/Vincristine in Relapsed/Refractory Aggressive B-cell Lymphoma

Abstract

Purpose:

The aurora A kinase inhibitor alisertib demonstrated single-agent clinical activity and preclinical synergy with vincristine/rituximab in B-cell non-Hodgkin lymphoma (B-NHL). This phase I study aimed to determine the safety and recommended phase II dose (RP2D) of alisertib in combination with rituximab ± vincristine in patients with relapsed/refractory aggressive B-NHL.

Patients and Methods:

Patients with relapsed/refractory, diffuse, large, or other aggressive B-NHL received oral alisertib 50 mg b.i.d. days 1 to 7, plus i.v. rituximab 375 mg/m² on day 1, for up to eight 21-day cycles (MR). Patients in subsequent cohorts (3 + 3 design) received increasing doses of alisertib (30 mg starting dose; 10 mg increments) b.i.d. days 1 to 7 plus rituximab and vincristine [1.4 mg/m² (maximum 2 mg) days 1, 8] for 8 cycles (MRV). Patients benefiting could continue single-agent alisertib beyond 8 cycles. Cell-of-origin and MYC/BCL2 IHC was performed on available archival tissue.

Results:

Forty-five patients participated. The alisertib RP2D for MR was 50 mg b.i.d. For MRV (n = 32), the RP2D was determined as 40 mg b.i.d. [1 dose-limiting toxicity (DLT) at 40 mg; 2 DLTs at 50 mg]. Drug-related adverse events were reported in 89% of patients, the most common was neutropenia (47%). Seven patients had complete responses (CR), 7 had partial responses (PRs); 9 of 20 (45%) patients at the MRV RP2D responded (4 CRs, 5 PRs), all with non-germinal center B-cell (GCB) diffuse large B-cell lymphoma (DLBCL).

Conclusions:

The combination of alisertib 50 mg b.i.d. plus rituximab or alisertib 40 mg b.i.d. plus rituximab and vincristine was well tolerated and demonstrated activity in non-GCB DLBCL.

ClinicalTrials.gov identifier: NCT01397825

Introduction

Non-Hodgkin lymphoma (NHL) is the sixth most common malignancy, accounting for 5% of all new cancer cases (1). Diffuse large B-cell lymphoma (DLBCL), an aggressive lymphoma, is the most common histologic subtype, accounting for 27% of new NHL diagnoses (2). DLBCL can be distinguished into cell-of-origin (COO) subtypes based on gene-expression patterns reminiscent of germinal center B-cell (GCB type) and non-GCB-derived disease subtypes (3, 4). Patients with COO identified as GCB-derived disease have better prognosis compared with patients with non-GCB disease (4). Patients with aggressive forms of NHL, such as DLBCL and transformed follicular lymphoma (TFL), often have a poor prognosis and relapse after first-line treatment with rituximab plus anthracycline-based chemotherapy such as CHOP (cyclophosphamide, doxorubicin, vincristine, and prednisone) or second-line high-dose chemotherapy with autologous stem cell rescue (5, 6). Other types of NHL, such as mantle cell lymphoma (MCL), have had several agents recently approved for use, but despite these new agents, most patients eventually succumb to the disease (7). Thus, there is an unmet clinical need for additional treatment strategies in these patients with such biological heterogeneity.

The search for novel targets in aggressive B-cell NHL has resulted in the study of the aurora kinases (8). Aurora A kinase (AAK) is a key mitotic regulator essential for centrosome maturation and separation, spindle assembly, chromosome alignment, and cytokinesis (9–11). Overexpression of AAK in experimental models results in transformation of normal cells, suggesting that this enzyme may be oncogenic (12). Indeed, AAK overexpression and/or amplification has been reported in lymphoma, leukemia, and in several solid tumor types (13–17) and is associated with tumor progression and poor patient outcomes (11, 18). AAK inhibition has been shown to lead to abnormal spindle formation, polyploidy, and subsequent cell death (19). Given the obligatory role of mitosis in tumor proliferation, an AAK inhibitor may have potential applications across a broad range of human tumors, making AAK a rational target for anticancer therapy.

Alisertib (MLN8237) is an investigational, oral, selective, small-molecule inhibitor of AAK with preclinical activity against a broad range of tumor types (20). Pharmacodynamic markers that support evidence for AAK inhibition by alisertib include tumor specimens with exposure-related increases in numbers of mitotic cells with spindle and chromosomal abnormalities (21) and in vivo tumor regression in lymphoma models (20). Alisertib has also shown preliminary antitumor activity with manageable toxicity in patients with hematologic malignancies (22–25).

The combination of AAK inhibition and microtubule disruption with vincristine has demonstrated synergistic antilymphoma activity in mouse models; the addition of the CD20 monoclonal antibody, rituximab, to the combination increased the activity of these two agents further (26, 27). Based on this encouraging preclinical activity, this phase I study was conducted to determine the safety and recommended phase II dose (RP2D) of alisertib in combination with rituximab ± vincristine.

Patients and Methods

Patients

Adults diagnosed with histologically confirmed DLBCL/TFL, MCL, or Burkitt lymphoma (with neoplastic cells expressing CD20), who had relapsed/refractory disease after 1 to 4 prior systemic therapies, and had relapsed following autologous stem cell transplantation (ASCT) or were not eligible for ASCT, were enrolled. Patients with transformed DLBCL/TFL from a previous indolent NHL or the concomitant presence of a component of low-grade lymphoma were not excluded from participation. Additional requirements included measurable disease (≥2 cm in the longest diameter), an Eastern Cooperative Oncology Group (ECOG) performance status of 0 to 2, and adequate bone marrow [absolute neutrophil count (ANC) ≥1,250 cells/mm³; platelet count ≥75,000/mm³], hepatic [total bilirubin ≤1.5 times upper limit of normal (ULN); serum alanine or aspartate aminotransferase (ALT/AST) ≤3 times ULN], and renal (calculated creatinine clearance ≥30 mL/minute) function. Patients were excluded if they had prior treatment with AAK inhibitors, diagnosis or treatment for a malignancy other than lymphoma within 2 years prior to first dose, evidence of active malignancy other than lymphoma, ASCT ≤3 months prior to enrollment or allogeneic stem cell transplantation at any time, or grade ≥2 peripheral neuropathy. All patients provided written informed consent.

Study design

This single-arm, open-label, multicenter phase I study (NCT01397825) was conducted at 10 investigational centers in the United States, with approval from institutional review boards/ethics committees in accordance with ethical principles founded in the Declaration of Helsinki including International Conference on Harmonization, Good Clinical Practice regulations and guidelines, and all applicable local regulations.

The study was conducted in two parts: a safety lead-in (part 1) evaluation of the combination of alisertib plus rituximab (MR) and a dose-escalation phase (part 2) evaluating the three-drug combination of alisertib plus rituximab and vincristine (MRV).

The alisertib starting dose and schedule for part 1 was chosen according to earlier single-agent alisertib studies (21, 24, 28) and low potential for drug-drug interaction between alisertib and rituximab. Patient enrollment to part 1 followed a modified 3 + 3 design that involved safety review of the first 2 patients who completed cycle 1 prior to enrollment of additional patients. Alisertib 50 mg was administered twice daily (b.i.d.) on days 1to 7 plus rituximab 375 mg/m² (intravenously, i.v.) on day 1 in 21-day cycles. Rituximab was administered for up to a maximum of 8 cycles but patients could continue to receive single-agent alisertib beyond cycle 8 until disease progression or unacceptable toxicity. The RP2D of the MR combination was identified when ≤1 of 6 patients experienced a dose-limiting toxicity (DLT) in cycle 1. If DLTs were observed in ≥2 patients at the alisertib 50 mg b.i.d. dose level, the alisertib dose would be deescalated.

MRV dose-escalation commenced with an alisertib dose approximately 50% of the RP2D identified for the MR combination (25 mg adjusted to a starting dose of 30 mg to accommodate the lowest available strength of 10 mg enteric-coated tablet). Alisertib was orally administered b.i.d. on days 1 to 7 plus rituximab 375 mg/m² i.v. on day 1 and vincristine 1.4 mg/m² (maximum 2 mg) i.v. on days 1 and 8 (drugs administered in the sequence described). Rituximab and vincristine were administered for a maximum of eight 21-day cycles, but patients could continue to receive single-agent alisertib beyond cycle 8 until disease progression or unacceptable toxicity. The alisertib dose was increased by 10 mg with each subsequent cohort in a 3 + 3 escalation design; the alisertib dose for the MRV triplet could not exceed the RP2D for the MR regimen. Cohort expansion was permitted for logistical reasons upon review and agreement by the sponsor and investigators.

At any time during study participation, if an individual in the MR or MRV cohort developed unacceptable toxicity to rituximab, treatment continued with the remaining agent(s) as tolerated. If a patient in the MRV cohort developed toxicity to vincristine, dose reductions according to the prescribing information were followed: the day 8 vincristine dose was omitted, or vincristine stopped completely prior to cycle 8 depending upon the toxicity and severity. If unacceptable toxicity to alisertib developed in patients receiving the MR or MRV combination, the patient was withdrawn from the study. Dose modifications were permitted for toxicities for alisertib and vincristine, with infusion adjustments possible for rituximab. Reescalations were not permitted once a dose reduction had been made for toxicity. Prophylactic growth factors were not mandated by the protocol.

Objectives and assessments

The primary objectives were to evaluate the safety and tolerability of the MR and MRV combinations, and to determine the RP2D of alisertib in both combinations, in patients with relapsed or refractory DLBCL and aggressive B-cell lymphomas. Secondary objectives included characterization of alisertib pharmacokinetics (PK) when coadministered with rituximab only (MR cohort) or when coadministered with rituximab and vincristine (MRV cohort); characterization of the effect of coadministered alisertib on vincristine PK in the setting of coadministration of rituximab; and antitumor activity in terms of overall response rate (ORR).

Toxicity was evaluated according to the National Cancer Institute Common Terminology Criteria for Adverse Events, version 4.03. DLTs were defined as grade 4 neutropenia lasting for >7 days (ANC <500 cells/mm³); grade 4 neutropenia with fever (oral temperature ≥38.5°C); grade 4 thrombocytopenia (platelets <25,000 cells/mm³) lasting for >7 days; platelet count <10,000 cells/mm³; grade 3 thrombocytopenia with clinically significant bleeding; any other grade ≥3 nonhematologic toxicity (with the exception of grade ≥3 nausea or diarrhea that occurred in the absence of optimal supportive care measures or brief [<1 week] grade 3 fatigue); and delay in initiation of the subsequent cycle of therapy >14 days due to treatment-related toxicity.

For alisertib PK investigations, blood was collected at prespecified time points following the morning alisertib dose on days 1 and 7 of cycle 1, and at the time of the alisertib morning dose on day 8 of cycles 2 to 8 in both parts of the study. At least 12 PK-evaluable patients were required at the RP2D for each of the treatment combinations, MR and MRV. Alisertib PK parameters calculated included maximum observed plasma concentration (C_max), first time to reach C_max (t_max), and area under the plasma concentration-time curve during a dosing interval (AUC_τ), where τ was the length of the dosing interval. To characterize vincristine PK in the MRV cohorts, blood was collected at prespecified time points following the vincristine dose on day 1 of cycles 1 and 2.

The alisertib dose was held in cycle 2 on days 1 to 3 to permit PK characterization of vincristine in the absence of coadministered alisertib. Vincristine PK parameters included C_max, AUC from time 0 to the time of the last quantifiable concentration (AUC_τ), AUC from time 0 to infinity (AUC_∞), and terminal elimination half-life (T_1/2z) for vincristine coadministered with rituximab alone (samples collected in cycle 2, days 1–3) and when coadministered with rituximab and alisertib (cycle 1, day 1).

Disease response and progression were determined by the investigators using the International Working Group criteria (29). Efficacy parameters included ORR [complete response (CR) + partial response (PR)], duration of response (DOR), and progression-free survival (PFS). Imaging was performed at the end of every other cycle (between days 14 and 21) starting at the end of cycle 2.

Tumor tissue was collected as an optional sample. Where COO disease subtype was not provided by the investigator and a sample was available, hematopathologists at the University of Arizona, Tucson, analyzed the samples; the hematopathologists were blinded to patient cohort and outcomes. To evaluate COO subtype, IHC was performed on 4-μm formalin-fixed paraffin-embedded (FFPE) sections using the automated Ventana Benchmark XT instrument (Ventana Medical Systems/Roche). Antigen retrieval was achieved by using a cell conditioning solution (CC1, Ventana Medical Systems), and antibody detection was visualized with the ultraView Universal DAB Detection Kit (Ventana Medical Systems). Rabbit anti-CD3 (2GV6), anti-CD10 (SP 67), anti-bcl2 (SP66), MUM1 (MRQ-43), anti-Ki67 (30–9), and anti-cMyc (Y69) monoclonal antibodies, and mouse anti-CD20 (L26), anti-bcl6 (GI191E/A8; Ventana Medical Systems), and anti-CD30 (Ber-H2; Dako) monoclonal antibodies were used. Hematoxylin and eosin-stained sections from FFPE tissue blocks and CD20 staining for each tumor were used to identify tumor areas and the presence of malignant B cells. The cutoff for tumor positivity was set at >30% of cells staining for CD10, Bcl6, and MUM1 staining (30). The tumor area with the highest percentage of tumor cell staining was used for COO analysis.

The COO was determined using the Hans algorithm (30). The expression of cMyc, Bcl2, and Ki67 was analyzed in the tumor cells with tumor positivity thresholds of >40% for cMyc (31) and >10% for all other antigens.

Analysis populations and statistical analyses

The safety population included all patients who received ≥1 dose of alisertib. The DLT-evaluable population included patients with cycle 1 data in the MR cohort and in the MRV dose-escalation cohorts, prior to enrolling at least 12 PK-evaluable patients at the RP2D for each of the MR and the MRV treatment combinations. The PK-evaluable population included patients for whom there were sufficient dosing and drug concentration-time data to permit noncompartmental PK analysis. The response-evaluable population included patients with measurable disease who received ≥1 dose of alisertib and had at least one post-baseline response assessment; this population was used for efficacy analyzes. Of note, the protocol was originally to proceed with a phase II portion following an interim analysis with a go/no-go decision for the first 25 evaluable patients enrolled at the MRV RP2D; if 8 or more patients achieved a response, the study would continue. In 2013, a sponsor decision was taken to not initiate the phase II portion of the trial, which was based on the sponsor’s reprioritization and not on any clinical or safety outcomes observed.

Statistical analysis was primarily descriptive and graphical in nature. Time-to-event data were analyzed by the Kaplan-Meier method. For alisertib PK, individual and mean plasma concentration-time data were plotted on days 1 and 7 of cycle 1 by dose level/cohort. For vincristine, individual and mean plasma concentrations were plotted over time for vincristine coadministered with rituximab and alisertib on day 1 of cycle 1, and vincristine coadministered with rituximab on day 1 of cycle 2, by alisertib dose level/cohort. Noncompartmental PK analysis using Phoenix WinNonlin version 6.2 was performed on individual concentration-time data.

Results

Patient characteristics

Forty-five patients were enrolled and started treatment across 10 sites in the United States between August 9, 2011, and January 14, 2014; 13 patients in part 1 and 32 patients in part 2. Baseline characteristics for patients in both parts of the study and overall are detailed in Table 1. The most common primary diagnosis was DLBCL (n = 35; 78%), followed by MCL (n = 6; 13%).

Table 1.

Patient demographics and baseline disease characteristics

	Part 1 MR (n = 13)	Part 2 MRV (n = 32)	Total (N = 45)
Median age, years (range)	61.0 (37–86)	65.5 (28–81)	64.0 (28–86)
Male, n (%)	10 (77)	19 (59)	29 (64)
Race, n (%)
White	12 (92)	28 (88)	40 (89)
Black	0	2 (6)	2 (4)
Asian	1 (8)	0	1 (2)
Other	0	1 (3)	1 (2)
Not reported	0	1 (3)	1 (2)
ECOG performance status, n (%)
0	7 (54)	15 (47)	22 (49)
1	4 (31)	13 (41)	17 (38)
2	2 (15)	4 (13)	6 (13)
Median international prognostic index score (range)	3 (1–4)	2 (0–4)	2 (0–4)
Ann Arbor stage, n (%)
I	1 (8)	2 (6)	3 (7)
II	0	5 (16)	5 (11)
III	3 (23)	11 (34)	14 (31)
IV	7 (54)	4 (13)	11 (24)
Other	2 (15)	5 (16)	7 (16)
Unknown	0	5 (16)	5 (11)
Disease type, n (%)
DLBCL not otherwise specified	8 (62)	27 (84)b	35 (78)
MCL	3 (23)	3 (9)	6 (13)
DLBCL associated with chronic inflammation	0	1 (3)b	1 (2)
B-cell lymphoplasmacytic lymphoma/immunocytoma	1 (8)	0	1 (2)
Othera	1 (8)	1 (3)	2 (4)
COO classification, n (%)
GCB	2 (15)	7 (22)	9 (20)
Non-GCB	8 (62)	12 (38)	20 (44)
Unknown or indeterminate	3 (23)	13 (40)	16 (36)
Median time from diagnosis, years (range)	1.5 (0.3–10.4)	1.7 (0.1–13.8)	1.6(0.1–13.8)
Evidence of extranodal involvement, n (%)
Yes	11 (85)	28 (88)	39 (87)
No	0	0	0
Unknown	2 (15)	4 (13)	6 (13)

Abbreviations: COO, cell of origin; DLBCL, diffuse large B-cell lymphoma; ECOG, Eastern Cooperative Oncology Group; GCB, germinal center B-cell; MR, alisertib plus rituximab; MRV, alisertib plus rituximab plus vincristine.

^aTFL to “double-hit” lymphoma and follicular B-cell lymphoma transformed to DLBCL.

^bTwo patients with DLBCL not otherwise specified and one with DLBCL with chronic inflammation violated protocol entry criteria as they were not CD20 positive. The 2 patients with DLBCL not otherwise specified discontinued in cycle 1, one with symptomatic deterioration and one with progressive disease; the patient with inflammation-associated disease withdrew consent for further treatment at cycle 4 after having stable disease.

Overall, 20 patients (44%) with DLBCL were identified as having non-GCB-derived disease subtype and 9 patients (20%) with GCB-derived disease subtype; the remaining 16 patients (36%) had unknown or indeterminate COO (Table 1). COO was documented by the investigator in 10 patients (22%) and determined by central analysis of tumor tissue sample at the University of Arizona, Tucson, in 19 patients (42%). Of the COO determined centrally, 19 (73%) of 26 available samples from DLBCL patients were evaluable. In 7 patients (27%), due to failure of CD20 staining, B cells could not be identified and thus COO could not be determined. Samples were classified as GCB (n = 3; 12%) or non-GCB subtypes (n = 16, 62%) according to the Hans algorithm. Representative cases of both GCB and non-GCB subtypes are shown in Supplementary Fig. S1.

Although all patients in the trial had received between 1 and 4 prior systemic therapies, with prior rituximab allowed, 6 patients (13%) had also relapsed from prior ASCT, and 17 (38%) patients had received prior radiotherapy. Five patients were confirmed by biopsy at baseline to have disease with bone marrow involvement.

DLTs and maximum tolerated dose (MTD) determination

In part 1, following a modified 3 + 3 design, 3 initial patients were enrolled to receive alisertib 50 mg plus rituximab 375mg/m². One patient experienced DLTs of grade 4 thrombocytopenia and grade 4 decreased neutrophil count for >7 consecutive days. A further 3 patients were enrolled at this dose level with no additional DLTs reported, and the RP2D was determined as alisertib 50 mg. An additional 7 patients were enrolled to the MR cohort to achieve a minimum of 12 PK-evaluable patients at this RP2D. In part 2, none of the 4 patients enrolled to the starting cohort of alisertib 30 mg b.i.d. plus rituximab 375 mg/m² and vincristine 1.4 mg/m² (maximum 2 mg) experienced a cycle 1 DLT. A fourth patient was added to the 30 mg b.i.d. cohort after the vincristine dose was reduced in 1 patient in error. Dose escalation proceeded to alisertib 40 mg; 1 patient experienced a cycle 1 DLT (grade 4 febrile neutropenia). No other cycle 1 DLTs were reported in 6 additional patients, and dose escalation proceeded to alisertib 50 mg; 2 of 3 patients experienced cycle 1 DLTs of grade 4 febrile neutropenia (n = 1), and grade 4 leukopenia, grade 4 stomatitis, grade 3 fatigue, and grade 3 enterobacter bacteremia (all in the same patient), so enrollment in this dose cohort was terminated. Alisertib 40 mg was determined to be the RP2D for the MRV regimen, and this cohort was expanded to provide further characterization of PK, safety, and clinical activity.

Treatment exposure and safety

All 45 patients were included in the safety population. Patients in the MR and MRV parts of the study received a median of 4 (range, 1–54) and 3 (range, 1–31) cycles of alisertib, respectively. Duration of treatment and primary reasons for study treatment discontinuation are summarized in Table 2. At the time of data cutoff (February 5, 2015), 5 patients remained on study treatment, 3 of whom had failed prior ASCT. After data cutoff, 1 patient (in the safety MR cohort) continued alisertib for a total of 79 cycles and then withdrew from treatment. Of the other 4 patients (all in the MRV RP2D cohort), one continued alisertib for 19 cycles (then discontinued for disease progression), one received 38 alisertib cycles [then developed myelodysplastic syndrome (MDS)], and 2 received 56 and 57 cycles, respectively, then continued on further alisertib treatment under their site’s single-patient Investigational New Drug program. The patient who developed MDS was a 73-year-old white male, ex-smoker, with non-GCB DLBCL and bone marrow involvement at study baseline (May 17, 2013), who had previously received rituximab-CHOP, ibritumomab tiuxetan, RICE (rituximab, ifosfamide, carboplatin, etposide), and ASCT with BEAM conditioning (carmustine, etoposide, cytarabine, melphalan), lymph node resection, and radiotherapy. A bone marrow biopsy on August 25, 2015, was considered consistent with a myeloid disease process without evidence for lymphoma involvement and MDS could not be excluded.

Table 2.

Treatment exposure and reasons for study treatment discontinuation

	Part 1 MR	Part 2 MRV
Alisertib dose, mg	50 b.i.d. (n = 13)	30 b.i.d. (n= 4)	40 b.i.d.a (n= 25)	50 b.i.d. (n= 3)	Total (n = 32)
Median cycles of alisertib, n (range)	4 (1–54)	2 (1–15)	4 (1–31)	3 (1–5)	3 (1–31)
Median duration of treatment, days (range)	70 (7–1,218)	28 (7–328)	69 (5–638)	77 (6–126)	62.5 (5–638)
atients receiving ≥8 cycles, n (%) Primary reason for study treatment discontinuation, n	3 (23)	1 (25)	8 (32)	0	9 (28)
PD	10	3	10	1	14
AE	0	1	4	1	6
Symptomatic deterioration	2	0	4	0	4
Voluntary withdrawal	0	0	3	1	4
Ongoing on study treatment, n	1	0	4	0	4

Abbreviations: AE, adverse event; b.i.d., twice daily; MR, alisertib plus rituximab; MRV, alisertib plus rituximab plus vincristine; PD, progressive disease; RP2D, recommended phase II dose.

^aRP2D.

All 45 patients experienced ≥1 AE, and 40 patients (89%) reported ≥1 drug-related AE. The most common drug-related AEs included neutropenia (47%), leukopenia (42%), anemia (38%), thrombocytopenia, diarrhea (each 33%), and fatigue (31%; Table 3). Thirty-seven patients (82%) reported ≥1 grade ≥3 AE; of these, 29 (64%) experienced drug-related grade ≥3 AEs, with the most frequent being neutropenia (42%), leukopenia (36%), anemia (22%), and thrombocytopenia (20%). Serious AEs (SAE) were experienced by 21 patients (47%); 11 patients (24%) reported drug-related SAEs. Drug-related SAEs reported in more than 1 patient were febrile neutropenia (n = 5, 11%), neutropenia, anemia, and stomatitis (each n = 2, 4%).

Table 3.

Most common drug-related AEs

Patients with events, n (%)	Part 1 MR (n= 13)	Part 2 MRV (n= 32)	Total (N= 45)
Any-grade drug-related AEs occurring in ≥10% of patients overall
Neutropenia	7 (54)	14 (44)	21 (47)
Leukopenia	8 (62)	11 (34)	19 (42)
Anemia	4 (31)	13 (41)	17 (38)
Diarrhea	4 (31)	11 (34)	15 (33)
Thrombocytopenia	7 (54)	8 (25)	15 (33)
Fatigue	3 (23)	11 (34)	14 (31)
Alopecia	3 (23)	6 (19)	9 (20)
Nausea	1 (8)	8 (25)	9 (20)
Stomatitis	2 (15)	7 (22)	9 (20)
Respiratory, thoracic, and mediastinal disorders SOC	2 (15)	5 (16)	7 (16)
Decreased appetite	0	6 (19)	6 (13)
Febrile neutropenia	0	6 (19)	6 (13)
Infections and infestations	1 (8)	5 (16)	6 (13)
Lymphopenia	2 (15)	4 (13)	6 (13)
Constipation	1 (8)	4 (13)	5 (11)
Decreased neutrophil count	1 (8)	4 (13)	5 (11)
Peripheral neuropathy	0	5 (16)	5 (11)
Musculoskeletal and connective tissue disorders SOC	1 (8)	4 (13)	5 (11)
Vomiting	1 (8)	4 (13)	5 (11)
Drug-related grade ≥3 AEs occurring in >1 patient
Neutropenia	7 (54)	12 (38)	19 (42)
Leukopenia	8 (62)	8 (25)	16 (36)
Anemia	4 (31)	6 (19)	10 (22)
Thrombocytopenia	3 (23)	6 (19)	9 (20)
Febrile neutropenia	0	6 (19)	6 (13)
Decreased neutrophil count	1 (8)	4 (13)	5 (11)
Lymphopenia	2 (15)	2 (6)	4 (9)
Stomatitis	0	4 (13)	4 (9)
Fatigue	1 (8)	2 (6)	3 (7)
Decreased platelet count	0	2 (6)	2 (4)
Palmar-plantar erythrodysesthesia syndrome	0	2 (6)	2 (4)

Abbreviations: AE, adverse event; MR, alisertib plus rituximab; MRV, alisertib plus rituximab plus vincristine; SOC, system organ class.

A total of 13 patients (29%) had dose reductions due to AEs (MR, n = 4; MRV, n = 9). Six patients (13%) discontinued study treatment due to AEs, including grade 4 spinal cord neoplasm (n = 1, MRV alisertib 30 mg); grade 5 respiratory failure, grade 1 pulmonary fibrosis, and grade 2 cough (n = 1 each, MRV alisertib 40 mg); grade 4 neutropenia, lobar pneumonia, and septic shock (n = 1, MRV alisertib 40 mg); and grade 4 thrombocytopenia (n = 1, MRV alisertib 50 mg). Of these AEs, the events of pulmonary fibrosis, neutropenia, and thrombocytopenia were considered by the investigator as related to study treatment. There were 4 onstudy deaths occurring within 28 days from the last dose and within 61 days of the first dose: respiratory failure (n = 2), septic shock (n = 1), and pleural effusion (n = 1), all of which were considered by the investigator to be attributable to underlying disease and unrelated to study treatment.

Pharmacokinetics

Following alisertib administration, t_max was achieved approximately 3 hours after dose with both MR and MRV. Alisertib exposures increased in an approximately dose-proportional manner over the range of 30 to 50 mg b.i.d., consistent with previous reports (32–34). The geometric mean AUC_τ of alisertib at 40 mg b.i.d. in the MRV combination at the day 7 assessment was comparable with that achieved during single-agent therapy (35), suggesting that coadministration of rituximab and vincristine had no clinically meaningful effect on alisertib PK. Importantly, the exposures of alisertib achieved at the RP2D of 40 mg b.i.d. in the MRV combination are within the previously characterized biologically active range for tumor AAK inhibition (33). Vincristine PK data were obtained (in at least one treatment cycle) from 32 patients, including 25 patients at the MTD/RP2D of MRV treatment. At the RP2D, AUC_∞ of vincristine when administered in the presence of alisertib was equivalent to that observed with vincristine in the absence of alisertib (geometric mean ratio, 97%; 90% CI, 82–115; n = 22; Table 4), indicating that coadministration of alisertib had no impact on vincristine PK.

Table 4.

Key alisertib plasma PK parameters in the safety lead-in cohort (part 1) and dose-escalation cohorts (part 2) and key vincristine PK plasma parameters (1.4 mg/m²; maximum 2 mg) in combination with rituximab (375 mg/m²) in the presence (C1D1) and absence (C2D1) of alisertib (30–50 mg)

	Alisertib dose, mg
	Part 1 MR		Part 2 MRV
	50 b.i.d.		30 b.i.d.		40 b.i.d.		50 b.i.d.
Alisertib plasma PK parameters	n= 13	n= 12	n= 4	n= 4	n= 25	n= 20	n= 3	n= 2
Cycle/day	C1D1	C1D7	C1D1	C1D7	C1D1	C1D7	C1D1	C1D7
Geometric mean Cmax, nmol/L (CV)	1,399 (57%)	2,587 (53%)	841 (39%)	1,665 (11%)	1,092 (34%)	1950 (55%)	1,667 (34%)	3,504
Median Tmax, hours (range)	3 (2–8)	2 (0–8)	3.5 (3–6)	3.1 (2–6)	3 (1–12)	3.1 (2–8)	6.1 (3–12)	2.5 (2–3)
Geometric mean AUCτ, nmol/L*h (CV)	8605a (66%)	18,667 (60%)	5,216 (51%)	11,849(29%)	7435b (41%)	14,925 (64%)	11,206 (57%)	31,784
Vincristine plasma PK parameters			n = 4	n = 4	n = 25	n = 17	n = 3	n = 2
Cycle/day	C1D1	C2D1	C1D1	C2D1	C1D1	C2D1	C1D1	C2D1
Geometric mean Cmax, ng/mL (CV)	-	-	120 (41%)	98.2 (47%)	82.9 (75%)	87.3 (127%)	149 (43%)	120c
Geometric mean AUCτ, ng*h/mL (CV)	-	-	69.2 (14%)	60.8c	60.1 (48%)e	67.8 (42%)	98 (42%)	68.4c
Geometric mean AUC∞, ng*h/mL (CV)	-	-	77 (14%)	69 (6.4%)d	78 (43%)f	77 (43%)g	111 (39%)	NR
Geometric mean CL, L/h (CV)	-	-	25.9 (14%)	29 (6.4%)	25.6 (50%)	25.9 (35%)	18 (39%)	NR
Mean t1/2z, hours (st. dev)	-	-	20.2 (5)	19.9 (0.9)d	20 (5.9)f	20.2 (4.7)g	25.6 (4.6)	NR

Abbreviations: AUC_τ, area under the plasma concentration–time curve during a dosing interval; b.i.d., dosed twice daily; C_max, maximum observed plasma concentration; CV, coefficient of variance; PK, pharmacokinetics; T_1/2z, terminal elimination half-life; T_max, first time to reach C_max.

^an = 12.

^bn = 22.

^cn = 2.

^dn = 3.

^en = 24.

^fn = 22.

^gn = 16.

Antitumor activity

Of 45 patients, 8 were excluded from the efficacy analysis due to a lack of post-baseline assessment. Among 37 response-evaluable patients, the ORR (CR + PR) was 38% (n = 14; Table 5). In part 1 (MR, n = 12), 2 patients (17%) achieved a CR and 1 patient (8%) achieved a PR. In part 2 (MRV, n = 25), 5 patients (20%) achieved a CR and 6 patients (24%) achieved a PR. Two of the responders were classified as having MCL; the remainder had DLBCL not otherwise specified. Nine of the 11 responders in part 2 were in the MRV RP2D cohort, including 1 patient with MCL who achieved a best response of CR. Table 6 lists classification of COO by investigator or central analysis for each individual patient with available COO data, with corresponding best response and treatment cycles administered. Of 29 DLBCL patients with COO data available, in those who were identified with non-GCB-derived disease (69%) 6 patients (30%) achieved CR, 3 patients (15%) achieved PR and 5 patients (25%) achieved SD as best response. None of the 7 response-evaluable patients with GCB-derived disease subtype achieved a response on study or received more than 5 cycles of treatment; this trend was not quite significant at the 0.05 level (P = 0.058, two-sided Fisher exact test). The time to first response ranged from end of cycle 1 to cycle 12, with 57% of patients achieving first response by the end of cycle 2. For all patients, median PFS was 5.1 months. In responding patients, median DOR was 10.2 months. Median DOR and PFS in patients with a best response of CR were 14.3 months (9.0–35.7) and 16.4 months (10.1–36.9), respectively.

Table 5.

Best overall response in the response-evaluable population

		Alisertib dose, mg
	Part 1 MR	Part 2 MRV				Total (N = 37)
n (%)	50 b.i.d. (n = 12)	30 b.i.d. (n = 3)	40 b.i.d.a (n = 20)	50 b.i.d. (n = 2)	Total (n = 25)
ORR (CR + PR)	3 (25)	1 (33)	9 (45)b	1 (50)	11 (44)	14 (38)
CR	2 (17)	1 (33)	4 (20)	0	5 (20)	7 (19)
PR	1 (8)	0	5 (25)	1 (50)	6 (24)	7 (19)
SD	5 (42)	0	6 (30)	1 (50)	7 (28)	12 (32)
PD	4 (33)	2 (67)	5 (25)	0	7 (28)	11 (30)

Abbreviations: b.i.d., twice daily; CR, complete response; MR, alisertib plus rituximab; MRV, alisertib plus rituximab plus vincristine; ORR, overall response rate; PD, progressive disease; PR, partial response; RP2D, recommended phase II dose; SD, stable disease.

^aRP2D.

^bThe original phase II interim go/no-go included a requirement of at least 8 responders in 25 RP2D response-evaluable patients.

Table 6.

Investigator assignment of COO or central analysis COO assignment with corresponding best response and treatment cycles administered, response-evaluable patients

Dosing cohort	COO	N	Best responses (# cycles received)
MR	GCB	2	SD (4), NA (1)a
	Non-GCB	8	2 CR (12, 79); 1 PR (6); 4 SD (4, 4, 4, 13); 1 PD (2)
MRV 30 mg	GCB	2	SD (5), NA (2)
	Non-GCB	2	CR (15), PD (2)
MRV 40 mg	GCB	5	3 SD (3, 4, 4b), 2 PD (2, 2)
	Non-GCB	9	3 CR (38, 56, 57), 2 PR (5, 17), 2 SD (4, 4), 2 PD (1, 2)
MRV 50 mg	Non-GCB	1	NA (1)

Abbreviations: COO, cell of origin; CR, complete response; DLBCL, diffuse large B-cell lymphoma; GCB, germinal center B-cell; LPL, lymphoplasmacytic lymphoma; MCL, mantle cell lymphoma; MR, alisertib plus rituximab; MRV, alisertib plus rituximab plus vincristine; PD, progressive disease; PR, partial response; SD, stable disease.

^aPatient with B-cell LPL (violated protocol entry criteria due to disease ineligibility).

^bPatient with DLBCL with chronic inflammation (violated protocol entry criteria due to disease ineligibility).

Immunophenotype and COO classification

For the 19 tumor tissue samples centrally analyzed for COO, Bcl2 was expressed in 2 of 3 tumors with the GCB subtype (67%) and in 9 of 16 tumors with the non-GCB subtype (56%); cMyc was expressed in no GCB subtype tumors and one (6%) non-GCB subtype tumor. One sample in the non-GCB group (6%) was positive for both Bcl2 and cMyc. The proliferation rate in two GCB samples (67%) was ≥80%, whereas in 4 (25%) of the non-GBC tumors, it was between 80% and 95%. Supplementary Fig. S2 shows the expression of Bcl2, cMyc, and Ki67, and the presence of T cells as depicted by CD3 staining, in the same samples as shown in Supplementary Fig. S1. Twenty-two response-evaluable patients were assessed for cMyc positivity by IHC. None of the 10 patients with an objective response had samples that were positive for cMyc expression using a cutoff of >40% positive cells. Similarly, no statistically significant difference in Ki67 positivity was found when comparing the responding and nonresponding patients.

Discussion

The addition of rituximab to initial chemotherapy has led to a significant improvement in outcomes for patients with DLBCL (36); however, there remains a substantial unmet need for patients who relapse, and especially those with the more aggressive non-GCB DLBCL subtype in whom outcomes are inferior to those with the GCB subtype (4, 37, 38). Several studies have shown clinical activity with single-agent alisertib in the treatment of lymphomas (24, 39). Preclinical work has shown that AAK inhibitors synergize with tubulin-disrupting agents, such as vincristine. Overexpression of AAK can override the mitotic spindle assembly checkpoint and lead to resistance to tubulin-disrupting agents (40). Indeed, combined inhibition of centromere assembly with alisertib, and inhibition of spindle assembly with vincristine, have been shown to have synergistic antilymphoma activity in preclinical models (26, 27).

Based on these promising data, the present phase I study evaluated, for the first time in a clinical setting, the safety and tolerability of alisertib in combination with rituximab (MR) or rituximab plus vincristine (MRV) in patients with relapsed/refractory aggressive B-cell NHL. We showed that alisertib 50 mg b.i.d. could be safely combined with rituximab (375 mg/m² administered every 3 weeks) with minimal need for dose modification. However, the addition to MRof vincristineon days 1and 8 of each cycle did potentiate the toxicity of the combination. The RP2D of MRV was determined to be alisertib 40 mg b.i.d. on days 1 to 7, plus rituximab 375 mg/m² on day 1 and vincristine 1.4 mg/m² (maximum 2 mg) on days 1 and 8.

AEs with MRV, including at the RP2D, were manageable with standard medical intervention, dose delay, and dose reduction as needed; notably, given the use of vincristine, only 16% of patients in the MRV group reported peripheral neuropathy. However, this may be associated with the dose-related nature of vincristine-inducted neuropathy (41) and the relatively limited treatment duration (median 3 cycles). The most commonly reported AEs were neutropenia, leukopenia, anemia, diarrhea, thrombocytopenia, and fatigue. This AE profile reflects AAK inhibition in proliferating cells and is consistent with that observed in phase II studies of single-agent alisertib therapy in relapsed/refractory multiple myeloma, NHL, chronic lymphocytic leukemia (24), and other tumor types (35). Five patients received extended treatment with single-agent alisertib; 1 patient in the safety MR cohort continued receiving alisertib for 79 cycles and 4 patients in the MRV RP2D cohort continued alisertib for 19, 38, 56, and 57 cycles. The patient continuing alisertib treatment for 38 cycles went onto develop MDS; although a temporal association exists with the alisertib dosing and development of MDS, the patient’s prior chronic bone marrow involvement with lymphoma and prior multiple chemotherapy regimens are confounding factors in determining a causal link.

Based on PK data from the safety lead-in and dose-escalation cohorts, alisertib PK in the MR and MRV cohorts was consistent with previous results for single-agent alisertib PK (33, 34), supporting the inference of no clinically relevant effect of rituximab and vincristine on alisertib PK over the alisertib dosage range of 30 to 50 mg b.i.d. Based on PK data from the dose-escalation cohorts, no clinically meaningful effect of alisertib on vincristine PK was observed over the alisertib dosage range of 30 to 50 mg b.i.d. in the presence of rituximab. At the MTD/RP2D of the MRV combination, AUC_∞ of vincristine in the presence of alisertib was equivalent to that observed with vincristine in the absence of alisertib. Individual values of vincristine AUC_τ and AUC_∞ in the presence of alisertib overlapped substantially with the range of individual vincristine AUC_τ and AUC_∞ observed in the absence of alisertib, suggesting that concomitant administration of alisertib did not produce a clinically relevant effect on vincristine PK at the MTD/RP2D. Taken together, these data support a lack of PK interactions between alisertib and vincristine.

Proliferation, as measured by Ki-67, is a negative predictive biomarker in both DLBCL and MCL, and targeting components involved in proliferation, such as AAK, is appealing in these diseases (13, 42, 43). In the present study, MRV was associated with promising clinical activity, consistent with preclinical studies (26). Efficacy analysis demonstrated that MRV had antitumor activity in patients with aggressive B-cell lymphomas, meeting the original phase II interim go/no-go requirement of at least 8 responders in 25 RP2D response-evaluable patients as 9 responses were achieved in 20 response-evaluable patients at the RP2D. Interestingly, the ORR for patients with non-GCB-derived disease (20/29) was 45%, but no patients with GCB-derived disease (9/29) achieved an objective response; however, it should be noted that COO was determined locally in approximately one third of cases, potentially affecting the consistency of COO analysis. Nevertheless, these results are important in the context of the typically inferior outcomes seen in patients with non-GCB-derived DLBCL (4, 38). A potential hypothesis for this apparently greater benefit of MRV in patients with non-GCB-derived DLBCL is that activated B-cell (ABC)-derived DLBCL is characterized by constitutive nuclear factor-kappa-B pathway (NF-κB) activation (44) and that there is cross-talk between the AAK and NF-κB pathways; as reported in epithelial ovarian cancer stem cells, which also have constitutive NF-κB activation, AAK inhibition resulted in cell-cycle arrest, NF-κB inhibition, and decreased cytokine production, thereby decreasing cell proliferation (45). AAK inhibition may thus be hypothesized to have specific activity in the setting of constitutive NF-κB activation. Additionally, investigation of other candidate biomarkers was an exploratory objective of the study; however, insufficient data were available to undertake any analyses.

Due to sponsor decision (not based on clinical or safety outcomes), the planned phase II dose expansion to 52 evaluable patients at the RP2D did not take place. However, the combination of alisertib plus rituximab and vincristine could be further developed by incorporating an inhibitor of Bruton tyrosine kinase, such as ibrutinib, to overcome the induction of aneuploidy and senescence cells (AASC) during therapy with alisertib (46). Indeed, preclinical studies in double-hit and double-expressor DLBCL have demonstrated that triple therapy with alisertib plus ibrutinib plus rituximab is synergistic and can inhibit AASCs induced by AAK inhibition, resulting in inhibition of proliferation and enhanced apoptosis (46). These data, together with the promising clinical activity of ibrutinib in both non-GCB DLBCL and MCL, form astrong rationale for anearly-phase clinical trial of ibrutinib/MVR in these diseases.

In conclusion, the data from this phase I study suggest alisertib in combination with rituximab plus vincristine has a manageable safety profile at the RP2D of 40 mg b.i.d. This study demonstrated that 9 of 20 patients with relapsed/refractory aggressive B-cell lymphomas treated at the MRV RP2D achieved objective responses. Alisertib is under further investigation in phase I lymphoma studies in combination with bortezomib and rituximab in patients with relapsed or refractory MCL/B-cell low-grade NHL (NCT01695941), and with romidepsin in patients with relapsed or refractory B-cell or T-cell lymphomas (NCT01897012).

Translational Relevance.

There is an urgent and unmet need to develop safe and effective treatments for relapsed/refractory B-cell non-Hodgkin lymphomas (B-NHL). Overexpression of aurora A kinase (AAK) frequently occurs in B-NHL, and AAK interference has demonstrated activity in multiple malignancies. Alisertib is an investigational AAK inhibitor that has shown single-agent preliminary activity in lymphoma. Preclinical studies have shown synergy and synthetic lethality between vincristine and AAK inhibitors in mantle cell lymphoma and diffuse large B-cell lymphoma mouse xenograft models, suggesting alisertib may have greater clinical activity when used in combination. Cell of origin (COO; GCB or non-GCB) was documented, or optional tumor tissue sample was provided for determination of COO, for 64% of patients, permitting exploration of whether gene-expression signatures were suggestive of signs of antitumor efficacy. Our findings support that alisertib plus rituximab and vincristine has a favorable safety profile, is well tolerated, and is associated with encouraging clinical activity in patients with relapsed/refractory non-GCB B-NHL.