Phase I Studies of Sirolimus Alone or in Combination with Pharmacokinetic Modulators in Advanced Cancer Patients

This article is featured in Highlights of This Issue, p. 4477

Sirolimus (Rapamycin) was discovered in 1975 and initially developed as an antibiotic because of its antiproliferative properties. The realization that this effect is exerted on bacteria and human cells in an indiscriminate manner led to preclinical and clinical studies of the drug as an immunomodulator and antineoplastic agent. Sirolimus was eventually approved by regulatory authorities for prophylaxis of renal allograft rejection and is now commercially available. Moreover, it has shown activity in some cancers, including Kaposi sarcoma and hepatocellular carcinoma, in addition to tolerability in combination with other targeted agents (1–5). However, its potential in oncology was not fully explored in favor of newly patented analogs that have since been approved in the treatment of renal cell and neuroendocrine carcinoma and have shown activity in other malignancies.

The mTOR is a serine/threonine kinase and a key regulator of protein translation via its ability to phosphorylate S6K and 4EBP1 (6–10). Sirolimus binds intracellularly to FK-506 binding protein 12 (FKBP12), inhibiting mTOR and resulting in growth inhibition in preclinical tumor models primarily by inducing cell-cycle arrest or apoptosis (11–13). Intermittent dosing schedules in animal models have shown antitumor activity equivalent to continuous administration, but without prolonged immunosuppression (13).

One putative obstacle to sirolimus development has been its low bioavailability, which is shared by its analog everolimus (14), and is not uncommon in oncologic drugs (15, 16). The pharmacokinetic profile of sirolimus has been well described in single-dose experiments in healthy volunteers and multiple-dose studies in renal transplant patients (17–19). Absorption after oral administration is rapid with a bioavailability of 14%, a volume of distribution of 12 L/kg (oral multiple dosing), and an elimination half-life of 63 hours (oral single dose). The agent is 92% protein bound and undergoes extensive hepatic and intestinal metabolism via CYP3A4 and CYP3A5, as well as excretion by P-glycoprotein. One potential strategy to increase bioavailability of rapamycin is coadministration with CYP3A inhibitors, such as ketoconazole or grapefruit juice, the latter exerting an effect predominantly on intestinal enzymes.

Importantly, an analog of sirolimus, temsirolimus, is a prodrug that is hepatically metabolized to the parent compound and has been approved in renal cell carcinoma at an intravenous dosage of 25 mg per week. On the basis of the reported temsirolimus pharmacokinetics and activity at this dose, we undertook 3 separate but simultaneously conducted phase I studies to find the dose of orally administered weekly sirolimus that would achieve exposure similar to weekly intravenous temsirolimus. We also aimed to evaluate the toxicity and pharmacokinetic profile of intermittently administered sirolimus in patients with advanced malignancies, assess the change in C_max and AUC when sirolimus was coadministered with CYP3A inhibitors (e.g., ketoconazole and grapefruit juice), and model the relationship between sirolimus whole blood levels and inhibition of the direct mTOR target p70S6 kinase. These results build on the prior published oncology experience with sirolimus on a daily schedule (20).

Patients with incurable cancer and no other effective therapy, older than 18 years of age, and with Karnofsky performance status 60% or more were eligible to participate. Written informed consent was obtained from each subject. All subjects required laboratory parameters that included leukocytes of >3,000/mL, absolute neutrophil count >1,500/mL, platelet count >100,000/mL, total bilirubin within institutional normal limits, serum SGOT and SGPT <2.5 times institutional upper normal limits, serum triglycerides <500 mg/dL, serum creatinine within normal institutional limits, and for the sirolimus alone study, a CD4 count >500/mL. Patients with severe immunodeficient states or uncontrolled brain metastases were excluded.

Between October 2004 and November 2008, subjects were enrolled in 3 separate Institutional Review Board–approved protocols at the University of Chicago (ClinicalTrials.gov Identifiers: NCT00707135, NCT00708591, NCT00375245)—sirolimus alone, sirolimus plus ketoconazole, and sirolimus plus grapefruit juice (Supplementary Table S1). Sirolimus was administered once weekly as a 1 mg/mL oral solution (sirolimus alone and sirolimus plus grapefruit juice studies) or as a 1-mg tablet (sirolimus plus ketoconazole study). All studies used adaptive escalation designs in which sirolimus dose levels were calculated with a goal of reaching an area under the exposure curve (AUC) equivalent to that observed with the approved 25 mg dose of temsirolimus (AUC 3810 ng-h/mL). At the completion of each cohort, the sirolimus AUC from the previous cohort was calculated, as well as the relative change compared with the previous cohort AUC. An estimate of the increase required in sirolimus dose to achieve the target AUC was made assuming a linear relationship between dose and AUC. Each dose escalation was intended to reach between 75% and 100% of the target dose. Due to gastrointestinal side effects experienced by subjects at the 60-mg sirolimus dose, all subsequent cohorts received sirolimus in divided doses-–initially doses were given 4 hours apart and then 24 hours apart in subsequent dose cohorts.

For the sirolimus plus ketoconazole study, sirolimus was administered in week 1, and ketoconazole was added to all subsequent doses at a fixed dose of 200 mg twice daily from week 2 on day 1, followed by 200 mg daily on the next 3 subsequent days. Because we were trying to inhibit CYP3A, we used the standard dosage approved for fungal infections and prophylaxis. For the sirolimus plus grapefruit juice study, sirolimus was administered alone in week 1 and with grapefruit juice starting in week 2, one day before sirolimus. Grapefruit juice (supplied by Florida Department of Citrus), 240 cc, was administered once daily without interruption. This dosing was based on research showing that the half-life of intestinal enzyme inhibition of grapefruit juice is 12 hours (21), thus providing time for modulation before sirolimus dosing.

A single cycle of treatment was defined as 4 consecutive weeks with no scheduled interruptions between treatment cycles. A minimum of 8 consecutive weeks (2 cycles) was administered before reevaluation of disease status.

Dose-limiting toxicities (DLT) were defined as the occurrence of any of the following that were deemed at least probably drug-related in cycle 1: grade 3 or higher nonhematologic toxicity except fatigue, nausea, vomiting; grade 4 thrombocytopenia or anemia; grade 4 neutropenia or grade 3 neutropenia associated with temperature higher than 38.3°C; inability to complete cycle 1 because of toxicity; or delay in administration of scheduled doses of sirolimus greater than 2 weeks because of drug-related toxicity of any grade. Both the sirolimus plus ketoconazole and grapefruit juice studies used a classic “3+3” dose escalation design. In the sirolimus-alone study, to obtain more samples for pharmacodynamic analysis, a “6+6” design was used, in which dose escalation occurred if 0 or 1 of 6 subjects at the current dose level experienced DLT. If 2 subjects experienced DLT, 6 more subjects were entered at that dose level. If a total of 2 or 3 of 12 subjects experienced DLT, escalation proceeded to the next dose level. If 4 or more of 12 subjects experienced DLT, then dose escalation was stopped, and the previous dose was declared the maximum tolerated dose (MTD). If 3 or more of the original 6 subjects at any dose level experienced DLT, dose escalation was stopped, and the next lower dose level was declared the MTD.

Sirolimus whole blood concentrations were determined as previously described at the University of Texas Medical School at Houston under the supervision of K.N.E (22). The concentration data were analyzed by a noncompartmental method because of limited number of patients for each dose level. The analysis was carried out using PK Solutions Software (version 2.0; Summit Research Services). In the sirolimus-alone study, blood was collected on day 1 (baseline, then 2 and 4 hours post dose), day 2 (before dosing, then 0.4 and 3 hours post dose), day 4 (48 hours post second dose), day 8 (before dosing), and day 29 (before dosing). In the sirolimus plus ketoconazole and grapefruit juice studies, blood was collected in week 1 before administration of sirolimus alone and then at 0.25, 0.5, 1, 2, 4, 8, 24, and 48 hours; in week 2 before administration of sirolimus and either ketoconazole or grapefruit juice and then at 0.25, 0.5, 1, 2, 4, 8, 24, and 48 hours; and in week 3 before sirolimus administration.

mTOR phosphorylates p70S6 kinase (p70S6K) at threonine 389 that most closely correlates with activity in vivo (23). p70S6K phosphorylation at Thr389 was measured in peripheral blood lymphocytes (PBL) as a potential biomarker of rapamycin activity. Only subjects in the sirolimus-alone study underwent determination of phosphorylated p70S6K. Blood for collection of PBL was drawn before sirolimus administration at baseline and analyzed again on week 1, day 2; week 1, day 4; week 2, day 1; and week 5, day 1. PBL were separated by Lymphoprep density centrifugation, followed by isolation of CD3⁺ cells using Human T cell Enrichment Kit (StemCell Technologies Inc., catalog # 14051A) The CD3⁺ cells were then treated with medium or PMA and ionomycin for 1 hour. The cells were lysed and Western blots for phospho- and total p70S6K were conducted. The results were analyzed using UN-SCAN-IT software and plotted using SigmaPlot. An indirect response model (24) with rebound effect (25) was used to describe our pharmacodynamic data and estimate the inhibitory effect of sirolimus. This model was built using NONMEM (version VII, level 1, ICON) in conjunction with a gfortran compiler. First-order conditional estimation (FOCE) with interaction and Advan 8 were applied.

Within-patient absolute changes in pharmacokinetic parameters of sirolimus before and after coadministration of ketoconazole or grapefruit juice were estimated. At each dose level, the signed rank test was used to determine whether these changes were significantly different from zero. The relationship between changes in pharmacokinetic parameters and dose level were assessed using the nonparametric trend test. The relationship between observed (DV) and predicted p70S6 kinase phosphorylation at Thr389 was evaluated using a linear regression model among all samples. Response was assessed using RECIST 1.0.

A total of 138 subjects were enrolled in the combined studies (Supplementary Table S1), and their characteristics are summarized in Supplementary Table S2.

Toxicity and DLT observed on all 3 studies are summarized in Tables 1 and 2, respectively. The most common toxicities attributed to sirolimus observed throughout all 3 studies were hyperglycemia, hyperlipidemia, lymphopenia, anemia, and diarrhea. In the sirolimus-alone study, the single weekly dose of 60 mg was associated with significant diarrhea and gastrointestinal upset, so subsequent cohorts were administered split doses, either 4 or 24 hours apart. Gastrointestinal toxicity on the split dose level consisted of diarrhea, nausea, and emesis. These were managed conservatively with antidiarrheals (loperamide or diphenoxylate/atropine as needed) and antiemetics. All subjects were able to continue therapy, although one required dose interruption because of diarrhea.

Of the total enrolled 138 patients, 101 patients were evaluable for pharmacokinetics (Tables 3 and 4). Sirolimus concentrations were significantly increased by both ketoconazole and grapefruit juice at all dose levels (Supplementary Table S3). The highest dose levels in both studies exceeded our minimal target AUC goal of 3810 ng-h/mL. When sirolimus was administered alone, gastrointestinal toxicity necessitated splitting the dose into 2 equal administrations. Although 5 subjects were accrued to the 45 mg 24-hour apart dose level, only 2 subjects were evaluable for pharmacokinetic analysis. Data from this cohort suggested that a 90-mg total weekly dose would achieve AUC close to the target, and thus accrual to the sirolimus-alone study was stopped.

We chose to examine a phosphorylation site on the direct mTOR substrate, Thr389 on p70S6K, as a pharmacodynamic marker of sirolimus activity in the sirolimus-alone study. We observed inhibition of phospho-p70S6K 48 hours after administration of sirolimus at all dose levels (Fig. 1, P = 0.006). Interestingly, even at the 10-mg weekly dose, phospho-p70S6K activity was suppressed 1 week after the initial dose (data not shown).

In addition, we modeled the relationship between suppression of phospho-p70S6K and sirolimus concentration. In our model, the change of phospho-p70S6K was described by k_in (zero-order constant rates for gain) and k_out (first-order constant rate for loss) in which baseline phosphorylation is 100%. Sirolimus has an inhibitory effect on the k_in decreasing the level of phospho-p70S6K. The inhibition exhibits linear correlation with the concentration of sirolimus (C_siro) with a coefficient of k_effect. The observed and predicted phosphorylation levels were well correlated (Fig. 2A). The production rate of phospho-p70S6K decreases to k_in × (1 − k_effect × C_siro), whereas the patients are taking sirolimus. Paradoxically, in 6 of 12 subjects, we observed a rebound effect on phospho-p70S6K in which phosphorylated protein levels exceeded baseline at later time points (Fig. 2B). We determined that the rebound was driven by the concentration of sirolimus, expressed as |$k_{rebound} \times C^\prime _{siro}$| (⁠|$^\prime _{siro}$| is the concentration of sirolimus in the rebound compartment). The estimated phospho-p70S6K in addition to the rebound was our prediction for the patients with rebound effect. The typical value for k_in, k_out, k_effect, and k_rebound were 5.5, 0.05, 0.08 mL/ng, and 11.3 mL/ng, respectively.

In the collective studies, one partial response was observed. This subject, diagnosed with epithelioid hemangioendothelioma with hepatic metastases, was treated with sirolimus and grapefruit juice. She had been previously treated with sorafenib and developed progressive disease. She remains on sirolimus (with grapefruit juice) more than 3 years after enrollment. Stable disease was observed in 16 (40%), 16 (28%), and 11 (27%) subjects in the sirolimus alone, sirolimus plus ketoconazole, and sirolimus plus grapefruit juice studies, respectively.

These 3 studies confirm that oral administration of sirolimus is feasible in oncology patients, extending the work of Jimeno and colleagues (20). Our studies show that weekly oral sirolimus can achieve drug exposure (AUC) similar to that observed with its parenteral prodrug, temsirolimus, and the recommended phase II doses from this study are 90, 16, and 35 mg when administered alone, with ketoconazole, and with grapefruit juice, respectively. Notably, the target AUC was attainable at significantly lower sirolimus doses when combined with CYP3A inhibitors, either ketoconazole or grapefruit juice. In fact, when sirolimus was combined with grapefruit juice, target AUC was observed at 15, 25, and 35 mg. Toxicity observed with weekly sirolimus was typical of other mTOR inhibitors with glucose, lipid, and hematologic alterations being most common. In general, weekly sirolimus was well tolerated with relatively few serious adverse events or DLT as defined in these studies.

These findings carry important implications for development of mTOR inhibitors and oral antineoplastic agents. Sirolimus is commercially available with a safety record much longer than any of its analogs. Moreover, the patents for sirolimus for most oncology uses have expired, and thus the potential exists for substantial cost savings over its analogs. Therefore, sirolimus represents a viable cancer drug whose development would offer several advantages with further cost savings realized by combining the drug with agents that inhibit its metabolism. We hope that our phase I studies provide a basis for initiation of comparative effectiveness studies relative to the more expensive sirolimus analogs.

Many drugs contain warnings to avoid grapefruit juice or other members of the Rutaceae family, including Seville oranges and pummelo, but this is the first cancer study to harness this drug–food interaction. In 1989, grapefruit juice was serendipitously found to inhibit the metabolism of the calcium channel blocker, felodipine (26, 27). Further research discovered that grapefruit juice is a potent inhibitor of intestinal CYP3A4, CYP1A2, and CYP2A6 (27). In fact, small bowel enterocyte CYP3A4 protein levels begin to decrease within hours of grapefruit juice administration, and the effect is maximal if grapefruit juice is ingested simultaneously or within the previous 4 hours of drug administration. Furthermore, the half-life of the effect is approximately 12 hours and enzyme levels are reduced by a mean of 62% even after 5 days of grapefruit juice consumption (21). The reduction in protein level is likely posttranscriptional requiring de novo CYP3A4 synthesis (27). Although the magnitude of the interaction is highly variable, it is reproducible within individuals and seems to be dependent on small bowel CYP3A4 content, that is, persons with the highest intestinal CYP3A4 content have the largest reduction in enzyme levels and subsequently the greatest effect on drug metabolism (27–30).

Interestingly, different grapefruit juice formulations seem to vary in inhibitory potency and therefore, in consultation with the Florida Department of Citrus, we used a frozen concentrate product that was tested for furanocoumarin levels before delivery of each batch. This ensured consistency across cohorts and is something that must be kept in mind for future studies or applications. Of course, one advantage of grapefruit juice is that it is nontoxic without risk of overdose. Therefore, we have at our disposal an agent that can markedly increase drug bioavailability (in this study by approximately 350%) and, critically in the current environment, decrease prescription drug spending on many agents metabolized by P450 enzymes.

The optimal dose and schedule of sirolimus and its analogs are still being debated with current administration schedules varying from weekly (temsirolimus), every other week (ridaforolimus), or daily (everolimus). Daily dosing of sirolimus, the universal and approved schedule used in the organ allograft setting, has been explored in cancer patients and seems feasible (20). One could, therefore, reasonably expect success by administering sirolimus in the same indications as its analogs. On the basis of the relatively long half-life of sirolimus, our studies hypothesized that sirolimus could be administered on a weekly schedule that would, in turn, simplify administration and provide a period of immune cell recovery and reduce immunosuppressive effects. Weekly administration seems feasible and acceptable to patients.

In all subjects, a decrease in PBL phospho-p70S6K expression in response to a dose of sirolimus was observed at 48 hours as expected. However, at later time points in some subjects, PBL phospho-p70S6K expression increased above baseline, which was labeled a rebound effect. The mechanisms underlying this rebound are unclear as sirolimus dosing was continued as planned. Feedback loops clearly exist that can stimulate protein activation in response to mTOR inhibition and such a loop has been described linking p70S6K inhibition with activation of AKT through IRS1 (31). Whether this was operational in subjects enrolled in this study and why this was not observed in all subjects needs further investigation. At this time, the impact of the rebound in phospho-p70S6K on efficacy and immune function is unclear.

This work and that of others (32–34) would suggest that the optimal biologic dose of mTOR inhibitors might be lower than those currently being administered. This study showed target inhibition, albeit in peripheral blood, at an appreciably lower dose than the highest dose administered. In fact, lower doses of all the available mTOR inhibitors have not been fully explored for comparative effectiveness. Dose-finding studies are not necessarily profitable for manufacturers, especially if the drug has been approved, priced, and lower doses prove to be equally active. However, the societal benefits with respect to quality of life and health care costs could be substantial. Zidovudine in the treatment of human immunodeficiency virus provides a poignant example of the benefits associated with examination of lower doses (35, 36).

There are challenges to incorporating grapefruit juice, ketoconazole, or any inhibitor of drug metabolism into regular practice. For a nondrug product such as grapefruit juice one has to assure consistent potency. Moreover, interactions with other concomitantly administered medications must be examined cautiously as well as adherence to regimens that would presumably incorporate at least 2 agents. In a clinical trial setting, we were able to monitor and control these factors. However, there is now considerable evidence that sirolimus can be tolerably and feasibly administered to cancer patients (1–5, 20). There are ostensible and indubitable advantages to developing sirolimus over other mTOR inhibitors, whereas this work describes strategies to improve bioavailability by combining sirolimus with CYP3A inhibitors. Funding for this development should be critically contemplated by governmental and public agencies, as well as potentially nonprofit biomedical companies (37).

No potential conflicts of interest were disclosed.

Conception and design: E.E.W. Cohen, M. Kocherginsky, M.J. Ratain

Development of methodology: E.E.W. Cohen, M. Kocherginsky, K.N. Eaton, L. House, J. Ramirez, T.F. Gajewski, M.J. Ratain

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): E.E.W. Cohen, C. Hartford, K.N. Eaton, M.L. Maitland, K. Fox-Kay, K. Moshier, L. House, J. Ramirez, S.D. Undevia, G.F. Fleming, T.F. Gajewski

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): E.E.W. Cohen, K. Wu, C. Hartford, M. Kocherginsky, K.N. Eaton, T.F. Gajewski, M.J. Ratain

Writing, review, and/or revision of the manuscript: E.E.W. Cohen, K. Wu, C. Hartford, M. Kocherginsky, K.N. Eaton, M.L. Maitland, G.F. Fleming, T.F. Gajewski, M.J. Ratain

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): E.E.W. Cohen, Y. Zha, A. Nallari, K. Moshier, M.J. Ratain

Study supervision: E.E.W. Cohen, K. Moshier, M.J. Ratain

The authors thank Dan King (Florida Department of Citrus), Dominick's Supermarkets, Human Immunologic Monitoring Facility, Pharmacology Core of the University of Chicago Comprehensive Cancer Center.

This work was supported by NIH (R21CA112951 to E.E.W. Cohen and P30 CA14599 to the University of Chicago Comprehensive Cancer Center); Department of Medicine, University of Chicago Pilot Project Grant; and the William F. O'Connor Foundation.

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.