Randomized Trial of Oral Cyclophosphamide and Veliparib in High-Grade Serous Ovarian, Primary Peritoneal, or Fallopian Tube Cancers, or BRCA-Mutant Ovarian Cancer

PARP 1 and 2 enzymes regulate DNA damage repair and maintain genomic stability in cells. Inhibition of DNA repair by small-molecule PARP inhibitors potentiates DNA damage caused by cytotoxic chemotherapies, including cyclophosphamide (1–3). Inhibition of PARP activity in the presence of deleterious mutations in the BRCA gene, which is involved in the homologous recombination pathway of DNA damage repair, can result in tumor cell death through the process of synthetic lethality (4, 5). Clinical activity is observed with PARP inhibitors alone and in combination with cytotoxic chemotherapy in patients with breast or ovarian cancers carrying germline BRCA mutations (BRCA-mutant; refs. 6–8). Clinical responses have also been observed with PARP inhibitors in patients with high-grade serous ovarian cancer (HGSOC), a disease known to have a high incidence of DNA repair defects even in patients who do not carry germline BRCA mutations (9).

Low daily doses of oral cyclophosphamide (Cytoxan; Bristol-Myers Squibb Company) in combination with other agents have demonstrated clinical activity in lymphomas and multiple solid tumors (10–13). Our phase I study of oral cyclophosphamide in combination with veliparib was well tolerated and demonstrated activity in patients with BRCA-mutant tumors: 6 of 13 patients experienced a partial response (PR), and 3 additional patients had prolonged disease stabilization (14). Based on this promising activity, we conducted a multicenter, randomized phase II trial to compare the response rate (complete plus partial responses; CR+PR) of veliparib in combination with oral cyclophosphamide with that of oral cyclophosphamide alone in patients with pretreated BRCA-mutant ovarian cancer or in patients with pretreated HGSOC, primary peritoneal, or fallopian tube cancers. This trial was designed to estimate the relative contribution of PARP inhibition to the activity of this combination in patients with known BRCA mutations or in tumors known to have a high incidence of DNA repair defects (9). Secondary objectives were to evaluate archival tissue and blood samples for mutations in genes involved in DNA damage repair and determine poly(ADP-ribose) (PAR) levels in peripheral blood mononuclear cells (PBMC) and levels of phosphorylated histone H2AX (γH2AX), a marker of DNA damage response, in circulating tumor cells (CTC) before and during treatment (15, 16). Archival patient tumor samples were sequenced for 211 genes involved in DNA damage repair thought to possibly affect the therapeutic potential of both cyclophosphamide and PARP inhibitors. We also performed gene expression profiling to examine whether the expression of specific DNA repair genes might correlate with PARP mRNA levels, BRCA mutation status, or response to therapy.

Patients 18 years of age or older with histologically documented BRCA mutation–positive ovarian cancer [documented deleterious BRCA1/2 mutation or a BRCAPRO score (17) of ≥30%] were eligible to participate. Patients with primary peritoneal cancer, fallopian tube cancer, or HGSOC were also eligible to participate, regardless of BRCA mutation status. All patients were required to have received at least one line of standard therapy and have measurable disease. A Karnofsky performance status ≥70% and adequate liver, kidney, and marrow function defined as an absolute neutrophil count ≥1,500/μL, platelets ≥100,000/μL, total bilirubin ≤1.5× the upper limit of normal (ULN), aspartate aminotransferase and/or alanine aminotransferase <2.5× ULN, creatinine <1.5× ULN were also required. Prior exposure to PARP inhibitors or cyclophosphamide was allowed unless previously administered in combination.

Previous anticancer therapy or surgery must have been completed at least 4 weeks before enrollment. Patients with treated brain metastases stable for greater than 4 weeks off steroids were eligible. This trial was conducted under an NCI-sponsored IND with institutional review board approval at each participating site. Protocol design and conduct followed all applicable regulations, guidance, and local policies [ClinicalTrials.gov Identifier: NCT01306032].

This was an open-label, multicenter, randomized phase II study of the combination of veliparib and oral cyclophosphamide compared with oral cyclophosphamide alone in patients with pretreated primary peritoneal cancer, fallopian tube cancer, HGSOC, or BRCA-mutant ovarian cancer. Veliparib (ABT-888) was supplied by the Division of Cancer Treatment and Diagnosis, NCI, under a Collaborative Research and Development Agreement with AbbVie. Cyclophosphamide was obtained from commercial sources.

Oral cyclophosphamide was administered at 50 mg once daily, alone or with oral veliparib at 60 mg once daily throughout a 21-day cycle, the same combination regimen studied in our prior phase I trial (14). Patients were required to maintain a diary documenting when drugs were taken and any associated side effects. There were no restrictions on food consumption. Adverse events were graded according to NCI Common Toxicity Criteria version 4.0. Doses of both drugs were reduced for grade ≥2 nonhematologic and grade 4 hematologic toxicities. Nonhematologic toxicities were required to have resolved to ≤grade 1, and hematologic toxicities to ≤grade 2 (except lymphopenia) before continuing treatment. Radiographic evaluation was performed at baseline and every three cycles to assess tumor response based on the RECIST version 1.1 (18).

The trial was randomized and used a phase 2.5 design, intending to enroll 65 patients per arm to have 80% power to permit a 0.10 alpha level one-sided test to compare clinical responses of 35% for the combination therapy to 15% for single-agent cyclophosphamide (19). In addition, there was 80% power to perform a 0.10 alpha level one-sided test to compare 6-month progression-free survival (PFS) probabilities between the arms. The study had a provision for an early stopping rule: if approximately 50% of the intended patients (approximately 32–33 per arm) had responses evaluated and the response rate on the combination arm was less than that of single-agent cyclophosphamide, then accrual would end whenever this was determined. Patient data were analyzed with and without being stratified by known BRCA mutation status.

Formalin-fixed paraffin-embedded (FFPE)–archived tumor tissue samples were collected, and the tumor content was assessed from an hematoxylin and eosin (H&E)–stained 4-μm section of the specimen. If tumor content was found to be less than 70% of the total cellular content in the section, a manual macrodissection of the remaining tissue was performed to enrich for tumor cells (Fig. 1). DNA and RNA were extracted using Qiagen AllPrep DNA/RNA FFPE Kits. For the whole-exome capture sequence analysis, a total of 500 ng fragmented DNA for each sample was used to make a sequencing library by hybridization with Agilent SureSelectXT Human All Exon 50Mb capture baits, followed with sequencing on the Illumina HiSeq 2000 platform. Gene expression profiling was performed on the Affymetrix U133plus2 GeneChip (methods available in the Supplementary Data). Mutation and gene expression data were analyzed to identify any subset of patients benefiting from veliparib treatment using the cross-validated adaptive signature design approach (20). The same data were also interrogated with a multivariate penalized Cox proportional hazards model to investigate if any of the genes were associated with the hazard of disease progression in either the cyclophosphamide only or combination cohorts.

Whole blood for PBMC and CTC isolation and analysis was collected from patients enrolled at the NCI only. Specimens for CTC analysis were collected into 7.5-mL CellSave tubes (Veridex) at baseline (before administration of study drugs), 24 hours after dosing on cycle 1, day 1, before drug on cycle 2 day 1, and just before each restaging (every 3 cycles); levels of γH2AX were determined as previously described (16). Blood for PBMCs was collected into 8-mL Cell Prep tubes (Becton Dickinson) on cycle 1 day 1 at baseline and at 4 and 24 hours after drug, on cycle 2 day 1 before dosing and 4 hours after drug, and just before each restaging; PAR, a product of PARP, was measured as previously described (21). PBMC and CTC sampling were repeated after patient crossover.

Seventy-five patients were enrolled (Table 1); treatment was discontinued for 1 patient due to adverse events, 1 patient withdrew from the study, and 1 patient died before the end of the first cycle, leaving 72 patients evaluable for response (Table 1). Of these, 37 received cyclophosphamide alone and 35 the combination as their initial treatment regimen. Patients were heavily pretreated, all having received prior platinum and taxanes with the exception of patient #1071 who did not receive taxanes. Two patients had received prior PARP inhibitor therapy (niraparib, olaparib, veliparib), and 3 patients had received prior cyclophosphamide. No patient was eligible based on a BRCAPRO score alone.

Grade 2/3 leucopenia and lymphopenia were the most common adverse events experienced by patients receiving cyclophosphamide alone or in combination (Table 2); grade 4 lymphopenia and grade 4 thrombocytopenia were reported in two separate patients receiving the combination, necessitating dose reduction. There was a trend toward increased myelosuppression with the combination compared with single-agent cyclophosphamide; however, both treatment regimens were well tolerated and the toxicities were easily managed.

The addition of veliparib to cyclophosphamide did not improve the response rate over cyclophosphamide alone, and patient accrual ended early per the stopping rule defined in the protocol. Out of 70 total patients with responses reported, 1 patient in each arm (#1095 and #1088) had a CR. PR was seen in 6 patients in the cyclophosphamide-only arm [7/36 (19.4%) responses overall; 95% confidence interval (CI), 8.2%–36.0%], 3 patients in the combination arm [4/34 (11.8%) responses overall; 95% CI, 3.3%–27.5%], and 1 patient who crossed over to the combination arm after progressing on the cyclophosphamide-only arm. Four of the patients who responded on the cyclophosphamide-only arm had BRCA-mutant ovarian cancer (including the patient who had a CR), two had HGSOC, and one had fallopian tube cancer. Two of the patients who responded on the combination arm had HGSOC (including the patient who had a CR), one had BRCA-mutant ovarian cancer, and one had fallopian tube cancer.

In addition, 6 patients in the cyclophosphamide-only arm and 5 patients on the combination arm had stable disease (SD) for six or more cycles of treatment, as did 4 of the 29 patients who crossed over to the combination treatment. One patient on the cyclophosphamide-only arm (#1056) had prolonged disease stabilization, receiving more than 32 cycles of treatment. Exome analysis of the tumor samples from this patient revealed mutation in BRCA2. Two patients had prolonged clinical benefit on the combination treatment, one with fallopian tube cancer (#1093; BRCA status unknown) and one with BRCA1-mutant ovarian cancer (#1087) who initially progressed on the cyclophosphamide arm and subsequently crossed over. Both were continuing on treatment at the time of data analysis, each having received over 38 cycles (more than 2 years) of treatment. The numbers of cycles of initial and crossover treatment per patient are shown in Fig. 2A–C. There was no improvement in PFS with the addition of veliparib to cyclophosphamide (median 2.3 and 2.1 months for cyclophosphamide-alone and combination treatment, respectively; Fig. 2D; P = 0.68 by two-tailed log-rank test), nor did stratification by BRCA status (either reported or determined from sequencing archival tumor samples) reveal a subset with improved PFS (Supplementary Fig. S1).

PAR levels were determined in PBMC samples collected before and after coadministration of veliparib from 10 patients on the combination arm and 11 patients who had crossed over to the combination arm. Four hours after treatment, PAR levels were reduced by at least 48% compared with baseline in all samples (mean, 85%; SD, 13%) and were rebounding in most patients by 24 hours after treatment (Supplementary Fig. S2). Sufficient CTC counts (≥6 CTCs) were isolated in samples from 10 (5 from each treatment arm) of 23 patients analyzed; counts ranged from 6 to 24 per 7.5 mL whole blood (data not shown). An increase in γH2AX-positive CTCs was observed in the only patient for whom quantitative analysis could be performed (data not shown).

A panel of 211 genes selected for their involvement in DNA repair (Supplementary Table S1) were interrogated by whole-exome capture sequencing and gene expression profiling in tumor tissue from 55 patients (27 treated with cyclophosphamide alone and 28 with the combination). All 55 patients had deleterious mutations (i.e., nonsynonymous mutations at coding regions) in at least 4 (and up to 70) of the genes evaluated, with an average of 9.3 mutations per patient (Table 3). The most common mutations observed, by far, were in TP53, followed by BRCA1; our patients displayed proportionally higher frequencies of mutations in DNA repair genes such as BRCA1, BRCA2, and APC than are commonly found in ovarian serous carcinoma (Table 4). We observed no significant difference in either the total number of genes mutated or the mutational frequency of any particular gene between those patients who did or did not respond to treatment. Gene expression profiling over the 211 DNA repair genes was suggestive of two different populations within the 55 patients, but these populations did not align with BRCA mutational status, patient response, or any other characteristic that we could demonstrate (Supplementary Fig. S3 and Supplementary Excel file).

We used the cross-validated adaptive signature design approach (20) to analyze whether the mutation status or expression levels of the genes in our DNA repair panel could be used to identify a subset of patients who benefited from PARP inhibitor treatment. Although 22 genes with a P value less than 0.05 were identified (Supplementary Table S2), none of the genetic alterations were significantly associated with veliparib treatment benefit when adjusted for multiplicity to control for the false discovery rate. The selected variables were therefore not sufficient to build a reliable predictor to select patients who would benefit from PARP inhibitor therapy.

The combination of PARP inhibitors with cytotoxic chemotherapy has been poorly tolerated with enhanced myelosuppression limiting the doses of chemotherapy that can be safely administered (22). The combination of oral cyclophosphamide with veliparib, however, was well tolerated and could be safely administered on a chronic schedule providing uninterrupted PARP inhibition to the majority of patients (14); therefore, we decided to address the question of the relative contribution of PARP inhibition to the clinical activity of the combination by comparing veliparib with oral cyclophosphamide to oral cyclophosphamide alone in ovarian tumors carrying BRCA mutations or in gynecologic cancers known to have a high incidence of DNA repair defects (9). Even though oral cyclophosphamide has demonstrated activity in combination chemotherapy regimens in a variety of tumor types, this is the first trial to document the response rate, using current staging and response criteria, of oral cyclophosphamide alone in this patient population. Previously, the only report of single-agent cyclophosphamide activity in ovarian cancer was from a study conducted in 1965 in 17 patients (11 postoperative patients who had undergone incomplete resections and 6 patients with disease recurrence) that reported responses by physical examination and time to clinical progression or overall survival (23). As demonstrated in the current trial, oral cyclophosphamide is a well-tolerated oral regimen associated with responses and prolonged disease stabilization in this pretreated population.

Solid tumors carrying DNA repair defects, such as breast and ovarian tumors carrying BRCA mutations, demonstrate increased sensitivity to PARP inhibitors or DNA-damaging chemotherapies. In this trial, 9 of the 11 patients who responded to either cyclophosphamide alone or the combination had either HGSOC or known BRCA-mutant ovarian cancer. In previous work by investigators from the Cancer Genome Atlas, analysis of HGSOC identified defects in the homologous recombination DNA repair pathway in 51% of 316 patient samples analyzed (9). This finding could account for the responses observed in our patients. However, we did not demonstrate an increase in the response rate to oral cyclophosphamide with the addition of veliparib as the initial therapy in this trial, or with the addition of veliparib at the time of disease progression for patients initially treated with oral cyclophosphamide alone. The relative sensitivity of tumor cells carrying various defects in the homologous recombination pathway to PARP inhibition is not well characterized, and therefore it is not known whether including patients likely to carry other defects in the homologous recombination pathway (HGSOC) as well as patients carrying known deleterious mutations in BRCA1 or BRCA2 in a relatively small sample set may have affected the overall outcome of the trial. We did analyze the PFS of patients treated with the combination, stratifying by BRCA-mutant status. BRCA status from tumor exome analysis exhibited a slight trend toward an effect in patients who received the combination treatment (P = 0.22), indicating that it might potentially play a role in prognosis.

The lack of increased response rate with the addition of veliparib to cyclophosphamide could also be due to the dose of veliparib employed in our study, which was below the 250 to 400 mg twice-a-day doses used in recent trials such as the Gynecologic Oncology Group (GOG) trial of single-agent veliparib and others (24–28). The safety of higher doses of single-agent veliparib used in the GOG trial and other studies had not been established when the current trial was initiated. Higher doses of veliparib in combination with cyclophosphamide may have resulted in more responses; however, the doses of study drugs used in this trial were established as the MTD in our prior phase I trial (14), had been shown to inhibit PAR levels in tumors, and allowed safe, uninterrupted dosing over the entire period of trial participation for patients receiving veliparib. We observed inhibition of PAR levels in PBMCs after veliparib in this trial; however, there was some recovery of PAR levels by 24 hours, favoring twice-a-day administration for veliparib to have continuous suppression of PARP activity.

It is not yet known what relative or absolute level of PAR inhibition is necessary for clinical efficacy; 90% inhibition has been measured in previous clinical trials (29). Because of the difficulties inherent in collecting research biopsies in patients with ovarian and peritoneal cancers, we measured PAR levels in PBMCs as a potential surrogate for tumor tissue in patients receiving combination treatment. We did demonstrate a decrease in PBMC PAR levels by an average of 85% 4 hours after the first administration of veliparib; however, given the modest response rates observed we could not correlate PAR inhibition to clinical benefit. PARP inhibitors can function both by inhibiting the catalytic activity of PARP, resulting in persistent, unrepaired DNA single strand breaks, and by trapping PARP–DNA complexes, interfering with DNA replication (30–32). Although a potent catalytic inhibitor of PARP (33), in cell lines, veliparib causes less PARP trapping than some other PARP inhibitors at catalytically inactivating concentrations (32). The relative contribution of PARP–DNA trapping to the clinical activity of PARP inhibitors in combination with cytotoxic chemotherapy is not known. In addition, the role of PARP in modulating the activity of low-dose cyclophosphamide is postulated but not proven in the clinic.

Platinum sensitivity appears to be one of the determinants of response to PARP inhibitor therapy (34–36). In our trial, all patients had received prior platinum; however, we did not collect consistent data to determine the fraction of patients who had platinum-sensitive disease and how that correlated with clinical benefit on either arm. In view of our randomized trial design, we presume but cannot prove that the arms were evenly balanced with regard to the number of patients with platinum-sensitive versus -resistant disease.

We also performed an exploratory analysis of the mutation status and expression levels of 211 selected genes involved in DNA damage response. Widespread defects in pathways such as homologous recombination, nonhomologous end joining, mismatch repair, Fanconi anemia, and DNA replication were observed; however, the presence of these DNA repair defects did not predict for response to either cyclophosphamide or the combination of veliparib and cyclophosphamide, nor was the known BRCA status of the patients informative in determining the likelihood of patient response. The clinical significance of the genetic aberrations we observed and the optimal agent(s) and dosages for treatment based on those observations need to be further defined for the aberrations in the context of the disease histology.

Various mechanisms have been proposed for the antitumor activity of oral cyclophosphamide, including inhibition of cd4(+)25+ T regulatory cell function (37). However, the underlying mechanism(s) that account for the observed antitumor activity of low-dose oral cyclophosphamide in patients are not known. Even though the addition of a PARP inhibitor did not improve the response rate, this trial establishes the activity of low-dose cyclophosphamide, an oral, well-tolerated treatment, in patients with HGSOC, primary peritoneal and fallopian tube cancers, and BRCA-mutant ovarian cancer.

No potential conflicts of interest were disclosed.

The content of this publication does not necessarily reflect the views or policies of the Department of Health and Human Services, nor does mention of trade names, commercial products, or organizations imply endorsement by the U.S. Government.

Conception and design: S. Kummar, R.J. Morgan Jr, A.P. Chen, S.M. Steinberg, J.H. Doroshow

Development of methodology: S. Kummar, D.R. Gandara, R.J. Morgan Jr, J. Ji, C.-J. Lih, P.M. McGregor III, P.M. Williams, R.J. Kinders, J.H. Doroshow

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Kummar, A.M. Oza, G.F. Fleming, D.M. Sullivan, M.J. Naughton, M.A. Villalona-Calero, R.J. Morgan Jr, A.P. Chen, D.E. Allen, C.-J. Lih, P.M. Williams, R.J. Kinders, B.A. Conley, J.H. Doroshow

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Kummar, A.M. Oza, M.A. Villalona-Calero, P.M. Szabo, A. Youn, A.P. Chen, C.-J. Lih, M.G. Mehaffey, P.M. McGregor III, S.M. Steinberg, P.M. Williams, R.J. Kinders, R.M. Simon, J.H. Doroshow

Writing, review, and/or revision of the manuscript: S. Kummar, A.M. Oza, G.F. Fleming, D.M. Sullivan, D.R. Gandara, M.J. Naughton, M.A. Villalona-Calero, R.J. Morgan Jr, P.M. Szabo, A. Youn, A.P. Chen, J. Ji, D.E. Allen, C.-J. Lih, M.G. Mehaffey, P.M. McGregor III, S.M. Steinberg, P.M. Williams, R.J. Kinders, B.A. Conley, R.M. Simon, J.H. Doroshow

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S. Kummar, J. Ji, D.E. Allen, M.G. Mehaffey, W.D. Walsh, P.M. McGregor III, J.H. Doroshow

Study supervision: S. Kummar, A.M. Oza, A.P. Chen, J. Ji, J.H. Doroshow

The authors thank Drs. Andrea Regier Voth and Yvonne Evrard (Leidos Biomedical Research, Inc.) for editorial assistance in the preparation of this article.

This project has been funded in part with federal funds from the NCI, NIH, under contract numbers HHSN261200800001E and N01-CM-2011-00032, N01-CM-2011-00038, N01-CM-2011-00071, N01-CM-2011-00099, and N01-CM-2011-00100, and by the Intramural Research Program of the NIH, National Cancer Institute, Center for Cancer Research.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.