Introduction

Vorinostat, a histone deacetylase (HDAC) inhibitor, exerts its antitumor properties through cell cycle arrest, tubulin acetylation, anti-angiogenesis, gene expression modulation, Wnt signaling modulation, protein acetylation, epidermal growth factor receptor (EGFR) down-regulation, DNA damage, and among other pathways [1–12]. Vorinostat has demonstrated clinical activity in cutaneous T-cell lymphoma and mesothelioma but has no significant single agent activity in other solid tumors [13–21]. The lack of single agent activity in solid tumors did not decrease the enthusiasm in developing this agent in combination with cytotoxic agents, especially in view of its demonstrated pre-clinical additive or synergistic antitumor activity [12, 22–26]. Unfortunately, this preclinical synergy has not translated into clinically relevant activity in numerous clinical trials, including a randomized phase III trial of carboplatin and paclitaxel in the first-line treatment of non-small-cell lung cancer (NSCLC) [27–29].

We have previously investigated a combination of vorinostat and 5-FU in patients with refractory solid tumors [30, 31]. The rationale for these studies was based on the pre-clinically demonstrated, vorinostat-induced, thymidylate synthase down-regulation in colorectal cancer models as well as the demonstrated synergy between 5-FU and vorinostat in colorectal cancer xenografts [22]. Despite the lack of evidence of thymidylate synthase down-regulation on serial tumor biopsies, we noted interesting clinical activity in patients with refractory colorectal cancer [30, 31]. In a phase I study, 24 patients with chemotherapy-refractory metastatic colorectal cancer were treated with daily vorinostat on days 1, 2, and 3 combined with infusional 5-FU on days 2 and 3 and repeated in 2-week cycles [30]. The median progression-free survival (PFS) and overall survival (OS) in this cohort were 4.4 and 9.2 months, respectively. Prolonged disease stabilizations were noted at vorinostat doses of 800 mg/day × 3 and higher [30].

To better define the activity of vorinostat in combination with 5-FU in chemotherapy-refractory colorectal cancer, we conducted a prospective randomized phase II clinical trial in this population. We selected two doses of vorinostat (800 and 1,400 mg) in combination with a standard dose of infusional 5-FU. The 800-mg dose was the lowest dose associated with prolonged disease stabilization, while 1,400 mg was the highest tolerable dose in combination with 5-FU in our prior phase I study [30]. The goal of this phase II study was to investigate the efficacy and safety of these two doses in patients with chemotherapy-refractory colorectal cancer and to determine whether either arm would be candidate for development into a future randomized phase III study.

Patients and methods

The main endpoint of this prospective randomized phase II clinical trial was to determine the 2-month progression-free survival rate with two doses of vorinostat, in combination with 5-FU and leucovorin in patients with refractory colorectal cancer. Secondary endpoints included PFS, OS, and safety as evaluated by the National Cancer Institute Common Toxicity Criteria Adverse Event version 4 (NCI CTCAE v4). Other secondary endpoints included vorinostat and 5-FU population pharmacokinetics and serial QTc interval assessment on the high-dose vorinostat arm.

Patients’ criteria

Patients with advanced metastatic colorectal cancer who had failed standard chemotherapy were eligible for enrollment. To be eligible for study treatment, patients should have documented radiographic progression within 6 months from completion of oxaliplatin-based therapy, irinotecan-based therapy, and cetuximab-based therapy (in the event of a KRAS wild-type tumor). In addition, patients should have exhibited refractoriness to fluoropyrimidine therapy. Refractoriness to fluoropyrimidine was defined as radiographic progression within 4 weeks following a minimum of 6 weeks of 5-FU or capecitabine-based therapy. Patients had to be ≥18 years of age, have an ECOG performance status of 0–2, and have an acceptable organ function: white blood cell count ≥3,000/10−6 L, absolute neutrophil count ≥1,500/10−6 L, platelets ≥75,000/10−6 L, serum creatinine ≤1.5 × upper institutional normal level, total bilirubin ≤ upper institutional normal level, and serum aspartate aminotransferase (AST) and alanine aminotransferase (ALT) ≤3 × upper institutional normal in the absence of liver metastases and ≤5 × upper institutional normal in the setting of liver metastases. Patients could not have received any chemotherapy within 3 weeks from initiation of study treatment with the exception of nitrosureas or mitomycin C, which required a 6-week interval before study treatment. Patients with brain metastases, grade ≥2 neuropathy, grade ≥2 QTc prolongation, or other severe intercurrent illnesses were excluded. Patients who were HIV positive and taking anti-retroviral medicines were excluded because of potential drug–drug interactions. No other HDAC inhibitors (such as valproic acid) or other investigational agents were allowed while patients were on study. Prior vorinostat therapy was not allowed. Pregnant or lactating women were excluded from participation. All consenting patients having the potential of conceiving agreed to use double contraception during the study period. The study and consent form were approved by the Institutional Scientific and Review Committee and the Institutional Review Board before the study was activated. All patients provided signed informed consent before study entry. The study was IRB approved and conducted in accordance with the Good Clinical Practice Guidelines as issued by the International Conference on Harmonization and the Declaration of Helsinki.

Treatment plan and dose modifications

Patients satisfying the inclusion criteria were randomized to receive vorinostat at 800 or 1,400 mg, given orally once daily for 3 consecutive days in 2-week cycles. Vorinostat (Zolinza®) was supplied by Merck Sharp and Dohme Corp as 100-mg capsules. Patients received all 8 or 14 designated capsules at the same time during the day, preferably with a meal. Leucovorin (LV) was given intravenously at 400 mg/m2 over 2 h after day 2 dose of vorinostat and was followed by 5-FU intravenous bolus at 400 mg/m2 and 5-FU infusion at 2,400 mg/m2 over 46 h (days 2 and 3 of vorinostat). 5-FU and leucovorin were obtained commercially. Patients were pre-medicated with a 5-HT3 antagonist on day 2, prior to 5-FU/LV. Other anti-emetics were used per investigator discretion. Dose modifications were allowed for both vorinostat and 5-FU/LV based on toxicity. Grade ≥3 toxicities attributed to vorinostat (such as fatigue and nausea/vomiting) required a dose-reduction starting the subsequent cycle. The first and second dose-level reductions for the 800-mg vorinostat arm were 600 and 400 mg, respectively. Up to four dose-level reductions were allowed on the 1,400-mg arm: 1,100, 800, 600, and 400 mg. No dose-reductions below 400 mg of vorinostat were allowed. Grace ≥3 hematological and non-hematological toxicities attributed to 5-FU required a dose-reduction by one dose-level starting the subsequent cycle. The first dose-reduction consisted of a reduction in 5-FU bolus to 300 mg/m2 and 5-FU infusion to 2,000 mg/m2. The second dose-reduction resulted in elimination of the 5-FU bolus and further reduction in 5-FU infusion to 1,800 mg/m2. The third dose-reduction consisted of further reduction in 5-FU infusion to 1,600 mg/m2. No further dose-reductions in 5-FU were allowed.

Response assessment

Patients underwent a CT scan of the chest, abdomen, and pelvis within 3 weeks before the start of treatment. A CT scan was repeated every 8 weeks (4 cycles). Patients with progressive disease were taken off study. Response status was determined as per RECIST 1.1 criteria [32].The primary endpoint of the study was progression-free survival rate at 2 months from treatment initiation. PFS was defined as lack of progression or death as per RECIST 1.1.

Pharmacokinetics

Vorinostat pharmacokinetics (PK) were evaluated at 0 (pre-vorinostat), 0.5, 1, 2, 3, 4, 6, and 8 h after vorinostat, on day 2 of cycle 1. 5-FU steady-state levels were measured at 6 h after the start of 5-FU infusion on day 2 of cycle 1. Vorinostat PK and 5-FU levels were evaluated on each treatment arm in the first 10 and 20 patients, respectively. The methodology was previously detailed [31].

Non-compartmental analysis of vorinostat plasma PK parameters was performed using WinNonLin Version 5.3 software (Pharsight Corp, Mountain View, CA). The pharmacokinetic parameters for each vorinostat dose were summarized using mean ± SEM. The parameters estimated for vorinostat and its two major metabolites (glucuronide-conjugate and 4-anilino-4-oxobutanoic acid) were peak levels (C max), time to C max, and area under curve from 0 to 8 h (AUC0–8h); the elimination half-life (T 1/2) was determined for patients with at least 3 terminal elimination times points with correlation coefficient (r 2) > 0.80.

QTc assessment

Since the effect of vorinostat on QTc interval prolongation had not been previously evaluated at doses exceeding 800 mg [33], we measured the QTc interval in the first 10 patients on the 1,400 mg/day dose-level. QTc interval was measured on a 12-lead EKG before, at 2, 4, and 6 h after the first dose of 1,400-mg vorinostat. In the event 2 or more grade ≥3 QTc prolongations were seen, the high-dose vorinostat arm would have been terminated.

Statistical methods

To evaluate the efficacy of the two different doses of vorinostat in combination with 5-FU/LV as measured by 2-month progression-free survival (PFS), a two-stage Simon design was utilized. Since 5-FU re-exposure in the setting of prior 5-FU failure has been associated with a 2-month PFS rate of approximately 50%, we deemed that any of our study arms would be interesting for future investigation if evidence was observed that the PFS rate exceeds 50% at 2 months. The two-stage design consisted of 15 patients on the first stage and 28 patients on the second stage (in each arm). Continuation to the second stage occurred only if ≥9 out of 15 patients on the first stage were alive and progression free at 2 months. Otherwise, if the observed count was less than the threshold, the corresponding arm of treatment was deemed not interesting for further investigation and the enrollment on that arm was halted. An arm was considered efficacious after the second stage if 27 or more out of 43 patients were alive and progression free at 2 months. The decision rules are associated with a type I error rate of 5%. The sample size (43 patients per arm) was selected so to provide 80% power in detecting an increase of 20% points as compared to the historical control rate of 50%. Subjects were randomized to either of the two treatment arms in a 1:1 fashion using a permuted block randomization scheme. In order to maintain balance between the two arm with regard to LDH (≤upper normal limit vs. >upper normal limit) and ECOG performance status score (0, 1 vs. 2), randomization was further stratified by these factors.

Results

Study enrollment and patient demographics

The study was conducted at Roswell Park Cancer Institute, Buffalo, NY and was registered under ClinicalTrials.gov as NCT00942266. Study enrollment started in August 2009 and was completed in December of 2010. A total of 58 patients were enrolled on study, 43 patients on the low-dose vorinostat arm and 15 patients on the high-dose vorinostat arm. All patients had confirmed 5-FU refractory colorectal cancer as defined in the inclusion criteria, and all had progressed following all standard chemotherapy for colorectal cancer. About 60% of patients had an elevated LDH, a known negative prognostic factor in patients with advanced colorectal cancer. The patient demographics per arm are detailed in Table 1.

Table 1 Patients’ demographics
Full size table

Toxicity

Both treatment arms were well tolerated with most common grade ≥2 toxicities being fatigue, nausea, hand and foot syndrome, and diarrhea. Nausea was more prevalent on the 1,400-mg vorinostat dose-level, while other toxicities did not appear to be influenced by dosing. No grade ≥3 neutropenia or thrombocytopenia was recorded, reflecting the non-myelosuppressive nature of this regimen. Grade ≥2 toxicities for the 800- and 1,600-mg vorinostat arms are summarized in Tables 2 and 3, respectively.

Table 2 Non-hematological toxicities on the 800 mg vorinostat arm
Full size table
Table 3 Non-hematological toxicities on the 1,600-mg vorinostat arm
Full size table

Pharmacokinetic

Mean (± SEM) steady-state day 1 plasma 5-FU levels, measured 6 h after 5-FU infusion initiation, were 416.7 ± 30.5 and 433.5 ± 46.3 ng/mL for patients receiving 800 (N = 19) and 1,400 (N = 10) mg of SAHA, respectively (P = 0.76).

A summary of the pharmacokinetic parameters of vorinostat and its two major metabolites is shown in Table 4. No significant PK differences were observed between patients following oral administration of vorinostat at the 800- and 1,400-mg doses. Plasma C max and AUC0–8h values of the two inactive metabolites, vorinostat-glucuronide and 4-anilino-4-oxobutanoic acid, were 3–7 times higher than corresponding values for vorinostat, the parent drug (P < 0.0001, one-way ANOVA). These results were consistent with prior reports [34]. There was no difference in pharmacokinetics for the metabolites between both arms, suggesting no increased metabolism on the high-dose arm.

Table 4 PK parameters of vorinostat and its major metabolites after oral administration of 800- and 1,400-mg doses
Full size table

QTc assessment

The first ten patients enrolled at the 1,400-mg vorinostat dose-level underwent serial EKGs after their first dose of vorinostat. Four of the 10 patients had a baseline grade 1 QTc prolongation. No grade ≥3 QTc prolongations were recorded. Four patients had grade 2 QTc prolongations after vorinostat, 3 of whom had a baseline Grade 1 QTc prolongation.

Efficacy

High-dose (1,400 mg) vorinostat

Fifteen patients were treated on the first stage of the study. Stable disease was documented at 2 months in 8 patients while 7 patients progressed. Since the number of patients without progression at 2 months did not reach the threshold of 9 patients required for the second stage of the study, accrual on this arm was halted. The median PFS and OS were 2.9 (95% CI: 1.6–5.6) and 6.7 months (95% CI: 3.0 to not-reached), respectively.

Low-dose (800 mg) vorinostat

Fifteen patients were treated on the first stage of the study. One patient with pulmonary metastases had a documented partial response (PR), and eight patients had stable disease (SD) for a PFS rate of 9 out of 15, supporting progression into the second stage of the study. An additional 28 patients were enrolled on the second stage for a total of 43 patients. Only 23 out of 43 patients were progression free (1 PR, 22 SD) at 2 months for a PFS rate of 53%. The median PFS and OS of the overall population were 2.4 (95% CI: 1.8–3.6) and 6.5 months (95% CI: 4.8–7.8), respectively (Fig. 1). The PFS rate did not reach the pre-specified threshold of 27 out of 43 patients, and the combination was not deemed interesting enough for further evaluation. Interestingly, lung target lesions appeared to respond better to treatment than hepatic lesions on waterfall plots (Fig. 2). Since KRAS MT tumors are more likely to metastasize to the lungs [35], we performed a multivariate analysis to evaluate the impact of KRAS, LDH, gender, and performance status on PFS and OS. None of these variables resulted in a statistically significant impact on either efficacy outcomes. An elevated LDH resulted in a trend in worsening in PFS (Hazard Ratio of 1.63, P = 0.197). Similarly, the waterfall plots did not suggest a better likelihood of response based on KRAS status (Fig. 2).

Fig. 1
figure 1

a Progression-free survival curve for the 800-mg vorinostat dose-level. b Overall survival curve for the 800-mg vorinostat dose-level

Full size image
Fig. 2
figure 2

a Waterfall chart for hepatic lesions and by KRAS status on the 800-mg dose-level. b Waterfall chart for lung lesions and by KRAS on the 800-mg dose-level

Full size image

Discussion

Vorinostat is a potent inhibitor of class I and II HDAC that has proven clinical activity against cutaneous T-cell lymphomas [36].This agent has significant synergy with 5-FU in pre-clinical models [22]. It has been postulated that this synergy may be related to intratumor thymidylate synthase down-regulation [22, 37]. However, we have failed to document such down-regulation in colorectal metastases, even at doses exceeding 800 mg/day, despite interesting preliminary clinical activity [30, 31]. We therefore hypothesized that the combination of 5-FU and vorinostat may be an effective combination in colorectal cancer and that its associated activity may not be related to thymidylate synthase down-regulation.

In this phase II randomized clinical trial, we set to investigate whether the combination of relatively high doses of intermittent vorinostat overcomes resistance to 5-FU in refractory metastatic colorectal cancer. Unfortunately, we did not see significant signs of activity on either arms of the study. The 1,400-mg dose-level was closed after the first stage, while the 800-mg dose-level accrued to the pre-planned total of 43 patients. The 2-month PFS rate on both arms was 53%, and only 1 PR was observed among 58 total patients. The OS of 6.5 months compares favorably to a previously reported best supportive care median of 4.6 months [38]. However, this number is inferior to other studies incorporating 5-FU infusional therapy in patients who previously failed 5-FU treatment [39, 40].

The potential explanations for the negative outcome of this study are numerous. At the forefront is the possibility that HDAC inhibition does not improve clinical outcomes when added to cytotoxic therapy. This would not be inconsistent with the clinical body of evidence in other tumor types and with other cytotoxics [27]. Other possibilities include the patient selection on this study. All of our patients were radiographically refractory to fluoropyrimidine. The rationale for this selection was predominantly based on a large subgroup analysis from a prior phase I study and not on 5-FU refractory pre-clinical models. Since pre-clinical synergy was demonstrated in 5-FU-sensitive models, a more appropriate clinical investigation would have focused on non-5-FU-resistant colorectal cancer patients. Another potential explanation for the poor outcome is the high percentage of patients with elevated LDH, a known negative prognostic factor in patients with stage IV disease.

We have shown, for the first time, evidence of bioavailability saturation of vorinostat at doses exceeding 800 mg. The 800- and 1,400-mg dose-levels were associated with a median C max of 3.1 and 3.6 μM, and AUC0–8h (h μmol/L) of 12.8 and 14.9, respectively. The lack of dose-PK effect at vorinostat doses exceeding 800 mg is likely due to absorption saturation rather than due to vorinostat self-induced metabolism, especially in view of the similar metabolite pharmacokinetic profile on both arms. Despite achieving concentrations associated with antitumor activity (C max > 2 μM), the AUC profile would be considered inferior to the conditions associated with antitumor activity when combined with 5-FU in vitro [22, 41]. When considering concentrations and exposure times similar to what was achieved in our study, vorinostat did not enhance 5-FU antitumor activity in vitro [42]. It has also been recently shown that vorinostat intratumor concentrations are considerably lower than corresponding serum concentrations, stressing further the need for higher and more sustained vorinostat serum concentrations [43]. Such favorable pharmacokinetics can be achieved with an intravenous formulation of vorinostat (no longer available) but are unlikely to occur with the current oral formulation [44].

Finally, we explored the impact of vorinostat on lung metastases following the observation of prolonged clinical control in patients isolated lung metastases. While the waterfall charts do suggest increased activity in lung metastases in comparison with liver metastases, we are unable to confirm whether this is related to selective activity of vorinostat in this subgroup of patients versus differences in tumor kinetics in relationship to the site of metastases. Since HDAC inhibitors have been reported to have enhanced activity in the presence of KRAS mutation [45], and in view of the increased prevalence of KRAS mutations in lung metastases [35], we investigated the impact of KRAS status on outcome in our study. KRAS status did not significantly impact PFS or OS.

In conclusion, we have failed to show clinically relevant activity for the combination of vorinostat (800 or 1,400 mg) and 5-FU in patients with refractory colorectal cancer. Further clinical development of this combination is not warranted.