Phase I Pharmacokinetic Study of the Novel Antitumor Agent SR233377¹

SR233377³(Fig. 1) is a member of a novel class of cytotoxic antitumor agents, the 4-aminomethyl thioxanthenone derivatives (1, 2). The exact mechanism of action of this agent is unknown; however, some members of this family of compounds preferentially inhibit DNA synthesis, and some, including SR233377, inhibit mammalian topoisomerase type II (3).

Preclinically, in a disc diffusion in vitro assay, SR233377 was selective for mouse solid tumors [e.g., colon adenocarcinoma 38 (C-38), pancreatic adenocarcinoma 03 (Panc 03), and mammary adenocarcinoma 16/C (Mam 16/C)] when compared with normal cells (human fibroblast and mouse gastrointestinal epithelium) and leukemias (murine P388 and L1210; Ref. 1). In vivo activity was observed against several s.c. implanted tumors including Mam 16/C, Panc 03 (curative), C-38 (curative), colon adenocarcinoma 51/A, pancreatic ductal carcinoma 02, colon adenocarcinoma 26, and i.v. implanted P388 leukemia (1).

Preclinically, a dose-limiting toxicity complex was observed in rodents (mice and rats) with rapid i.v. injection that included respiratory depression, seizures, and death at doses ≥100 mg/m² (1). This toxicity complex could be avoided in rodents when the drug was administered using a 1-h infusion (1).

The preclinical toxicology studies using a 1-h infusion schedule identified myelotoxicity as dose-limiting in all three species tested(mice, rats, and dogs; Ref. 4). There was a decrease in the total leukocyte count, segmented neutrophils, lymphocytes,platelets, and red cell mass. The nadir occurred between days 5 and 8 with full recovery by day 29. In vitro myelotoxicity assays comparing the effects of SR233377 on the bone marrow of mice, dogs, and humans also identified toxicity on the marrow of all three species;however, the mouse marrow was most sensitive and human marrow the least sensitive to this compound (5). Therefore, on the basis of the preclinical myelotoxicity profile, it was anticipated that myelotoxicity would be dose-limiting. Preclinical pharmacokinetic studies performed in mice showed most of the measurable drug in plasma was bound to proteins (96–100%); and higher amounts of drug were measured in tissues (liver, kidney, spleen, lung, pancreas, heart,brain, and skeletal muscle) compared with that detected in plasma (6).

As a result of the spectrum of preclinical solid tumor activity and acceptable toxicity profile using a 1-h infusion, SR233377 was advanced to Phase I clinical trials. The initial portion of our trial used a 2-h infusion given on day one of a 28-day cycle, but it was amended to be a 24-h infusion when unacceptable cardiotoxicity occurred. The initial infusion duration for this trial was chosen with the intent to maximize the single-dose schedule, after observing acute lethality in mice treated with a rapid i.v. infusion that could be ameliorated with a 1-h infusion. This is a report of our Phase I findings.

Patients with histologically confirmed solid tumors refractory to conventional therapy or for which no effective therapy was known were candidates for entry into this study. Eligibility criteria included:projected survival time of ≥12 weeks; a Zubrod performance status of≤2; no chemotherapy or radiotherapy for at least 28 days prior to treatment initiation (42 days for mitomycin C or nitrosourea); age ≥18 years; adequate bone marrow (WBC count≥4,000/mm³ and platelets≥100,000/mm³); adequate hepatic (serum bilirubin≤1.5 mg/dl) and renal (serum creatinine ≤1.5 mg/dl) function;negative pregnancy test in women of childbearing potential; no other coexistent medical problems of sufficient magnitude to jeopardize full compliance with the study; and signed informed consent to participate in the study. The consent form followed institutional guidelines consistent with those of the Food and Drug Administration. Patients were informed of known potential toxicities. After cardiac toxicity was identified, the protocol and consent form were amended, alerting patients of this toxicity and excluding patients with prior history of arrhythmias, congenital or acquired prolongation of the corrected QT interval (QT_C) of the electrocardiogram, and patients on anti-arrhythmic drugs or drugs known to prolong the QT_C.

Pretreatment tests and measurements included a complete history and physical examination with measurement of height and weight. Laboratory studies included a complete blood count with differential, prothrombin time, partial thromboplastin time, electrolytes, protein, albumin,blood urea nitrogen, creatinine, lactic dehydrogenase, alkaline phosphatase, calcium, uric acid, total and direct bilirubin, serum alanine aminotransferase, and urinalysis. Chest X-ray,electrocardiogram, and pertinent radiographic studies for evaluable/measurable disease were also performed. Both minimally (<3 prior radiation courses and/or chemotherapy regimens) and heavily pretreated patients (≥3 prior radiation courses and/or chemotherapy regimens) were eligible for this study. After drug-induced ventricular arrhythmias were identified, subsequent patients underwent 24-h Holter monitoring and evaluation by a cardiologist prior to treatment. Duration of infusion in this cohort of patients was extended to 24 h. During infusion, patients had continuous cardiac monitoring. If the QT_C was noted to exceed 550 ms, drug infusion was stopped; cardiac monitoring continued until the QT_C returned to pretreatment values, and the patient was felt to not be at risk of a cardiovascular event.

SR233377 was supplied by Sanofi Pharmaceuticals (Malvern, PA). Each glass vial of drug was clear and contained 20 ml of a 2.5 mg/ml yellow solution in isotonic citrate buffer (pH 5.5). Drug was kept refrigerated (2–8°C) until time of use. The appropriate dose of SR233377 was drawn into a syringe directly from one or more vials and infused using an autosyringe pump (Baxter Healthcare Corp., Deerfield,IL). For the 24-h infusion, drug was prepared at 6-h intervals.

The starting treatment dose of SR233377 was one-tenth the mouse LD₁₀ (drug dose that is lethal to 10% of the animals treated), 33 mg/m² (4),infused over 2 h. Each patient’s first cycle was administered in the inpatient pharmacokinetic unit with subsequent cycles administered in the outpatient setting. The drug was administered every 28 days. The initial dose escalation was 2-fold to 66 mg/m². Subsequent escalations proceeded by 50% increases for four doses(through 335 mg/m²) and then 33% to the highest dose tested (445 mg/m²).

Only two patients were entered into the 445-mg/m²dose level. Neither patient received their total drug dose because of acute, life-threatening ventricular arrhythmias at 55 and 57 min into the infusion. The protocol was then amended to prolong the infusion to 24 h with a starting dose of 225 mg/m². Four patients were treated at this dose, and all four completed the 24-h infusion.

Interim history, physical examination, toxicity evaluation (National Cancer Institute Toxicity Criteria), hematology, and chemistry profiles were performed weekly. Tumor measurements were performed every 8 weeks or more frequently if disease progression was suspected.

Blood was immediately chilled after collection in tubes containing potassium oxalate and sodium fluoride. A pretreatment blood sample was collected from all patients who agreed to have pharmacokinetic sampling. The designated time points for blood collection from patients treated using the 2-h infusion schedule were at 1-h during the infusion, immediately after infusion completion, then at 10, 20, 30,and 45 min and 1, 1.5, 2, 4, 6, and 24 h after infusion completion. The designated time points for blood collection from patients treated using the 24-h infusion schedule were 2, 4, 6, 9, 12,and 22 h during the infusion, immediately after infusion completion, then at 0.5, 1, 2, 3, 4, 6, and 24 h after infusion completion. Plasma was prepared from chilled blood samples by centrifugation at 4°C within 30 min after collection and then frozen at −20°C. Samples were kept frozen until time of assay by HPLC.

The sample loading portion of the HPLC consisted of a Varian 9090 autosampler (Walnut Creek, CA), a Varian 5500 pump, and a 0.4 ×3-cm column containing C₃ pellicular (30–50 μ)packing (Alltech, Deerfield, IL), whereas the analytical portion consisted of a Varian 9012 pump, a 0.2 × 3-cm guard column containing C₁₈ pellicular packing (Alltech), a 0.46 × 25-cm Ultrasphere ODS 5 μm C₁₈packing (Beckman Instruments, Inc., San Ramon, CA), and an Applied BioSystems 783A detector (Foster City, CA), set at 256 nm. Switching the injection from the sample loading portion to the analytical portion of the HPLC was accomplished using a Rheodyne valve 7000P (Cotati, CA)with pneumatic activator and a Rheodyne 7163 solenoid valve. The guard column and the analytical column fitted with precolumn were connected in tandem via the switching apparatus.

HPLC grade acetonitrile, deionized water, methanol, and ammonium acetate were obtained from J. T. Baker, Inc. (Phillipsburg, NJ). Reagent grade glacial acetic acid was obtained from Mallinckrodt(Chesterfield, MO). SR233422, a structurally related compound, was dissolved in deionized water and used as internal standard. Plasma samples (250 μl) were spiked with internal standard (100 ng; 20 μl of a 5-μg/ml solution), and then deionized water (250 μl) was added. An aliquot (100 μl) of the mixture was injected directly onto the sample loading portion of the HPLC, which had been conditioned with mobile phase 1, acetonitrile:0.05 m ammonium acetate(10:90, v/v) buffered at pH 4.5 using a flow rate of 1 ml/min. The sample was switched after 0.8 min in to receive mobile phase 2,acetonitrile:0.05 m ammonium acetate (38:62, v/v) at a flow rate of 0.75 ml/min for loading onto the analytical portion of the system. The switching valve returned the guard column to mobile phase 1 after 1.2 min to condition it to receive the next sample. The run time after sample loading onto the analytical portion was 13.5 min. The retention times of SR233377 and the internal standard were approximately 7.8 and 10.5 min, respectively. The overall recovery of SR233377 and the internal standard was near 100 and 95%, respectively. The concentration of SR233377 in patient samples was quantitated based on standard ratios determined from the peak areas of standards containing varying known amounts of SR233377 (five concentrations) and a fixed concentration of internal standard. The standards were assayed simultaneously with each patient’s samples. The assay was linear over the SR233377 concentration range of 10–1000 ng/ml, and 1:10 dilutions of plasma allowed extension of this to 10,000 ng/ml. The intra- and interassay coefficient of variation ranges were between 2.5 and 9.9%and 3.5 and 7.6%, respectfully.

Pharmacokinetic parameters were estimated by noncompartmental analysis using the MASTER_PK program resident within the RSI graphics package at Sanofi Pharmaceuticals, Inc. Actual times were used for all pharmacokinetic determinations. The peak plasma concentration for SR233377 was the highest drug concentration detected at or after the end of the infusion. The terminal phase half-life(t_1/2) was determined from the terminal rate constant, estimated by linear regression of the last portion of the plasma SR233377 concentration versus time profile using an unweighted fit. The area under the plasma concentration versus time profile(AUC_0-t) was calculated using the trapezoidal rule. AUC was extrapolated to infinity (AUC _0-inf, area under the concentration × time curve extrapolated to infinity) by dividing the last quantifiable plasma concentration by the terminal rate constant and summing this value with AUC_0-t. Clearance (Cl, drug clearance from plasma) was estimated according to the equation: Cl = D/AUC_0-inf,where D is drug dose.

The relationship of SR233377 peak plasma level to total dose and AUC_0-t to total dose was explored using regression analysis models (7). To better normalize the AUC_0-t distribution, a base 10 logarithmic transformation (log₁₀) was required. To explore nonlinear components in these relationships, polynomial regression models were fit, up to the third degree. R² (percentage of variability in the dependent variable accounted for by the regression model) values were used to compare the model adequacy among various polynomial regression models (8, 9). Detailed regression diagnostics were also examined. Predicted values from the log₁₀AUC_0-t model were transformed back to the original AUC_0-t scale to facilitate interpretation.

Patient characteristics are shown in Table 1. Twenty-four patients (16 men and 8 women) were entered into this study. The median age was 51 years (range, 25–82 years), and the median performance status was Zubrod 1 (range, 0–1). Eleven patients received prior chemotherapy only, one radiation only, 2 immunotherapy only, and 10 a combination of chemotherapy plus radiation. The majority of patients had intestinal cancer, but patients with other solid tumors (kidney, lung, melanoma, sarcoma, and thymoma) were enrolled.

The toxicity profile is shown in Table 2. There was minimal toxicity through 335 mg/m². One patient at 100 mg/m² and 2 patients at 225 mg/m² developed grade 1 mucositis. The mucositis was transient, lasting only a few days, most evident 7 days after therapy. Non-dose-limiting myelosuppression (maximum grade, 2) was the only other observable toxicity through 335 mg/m². The myelosuppression (manifested as mainly neutropenia) was noted 7–10 days after treatment and was completely resolved by day 14. Thrombocytopenia was not observed.

The dose-limiting toxicity with the 2-h infusion schedule was ventricular arrhythmias. In both patients treated at this dose (445 mg/m²), significant arrhythmias ensued. Cardiotoxicity had not been anticipated, and as a result, the initial patient (no. 19) did not have cardiac monitoring until after he had a transient loss of consciousness. The infusion was terminated immediately, and the patient regained consciousness within seconds. Blood pressure, heart and respiratory rate were recorded, and cardiac monitoring was begun immediately. The electrocardiogram showed ventricular tachycardia that lasted for approximately 5–7 min,followed by episodes of bigeminy and trigeminy. There was new onset of prolonged QT_C, which remained for ∼4 h. His systolic blood pressure increased by 20 mm Hg from baseline.

Because of confounding factors (patient anxiety, comedication with antihypertensive and antianxiolytics) of the initial event in patient no. 19, after much discussion patient no. 20 was treated at 445 mg/m² while on cardiac monitoring. During the initial 55 min of infusion, the patient was asymptomatic. However, 57 min into the infusion the cardiac monitor showed that she began having ventricular arrhythmia and within seconds later had a cardiac arrest during torsades de pointes and then had a seizure. The infusion was discontinued, and cardiopulmonary resuscitation was begun, but just prior to the use of electrocardioversion, the rhythm spontaneously converted to ventricular bigeminy and the patient awakened. The torsades de pointes lasted ∼2.5 min. After 2 h, the ventricular arrhythmia completely resolved. However, the patient had significant prolongation of her QT_C (maximum, 660 ms), which lasted for 6 h.

As a result of this acute, life-threatening toxicity, further treatment with the 2-h infusion of SR233377 ceased. After evaluation of cardiotoxicity in preclinical models and evaluation of patient data including pharmacokinetics, the trial was reopened using a starting dose of 225 mg/m² given over 24 h with the intention to stop the infusion at any point that the QT_C extended beyond 550 ms. Four patients were treated with the 24-h infusion, and continuous monitoring of QT_C occurred. All four patients completed their 24-h infusion (one to four courses), and a 10–16% increase in their QT_C over baseline was observed during the infusion. None of the increases were >550 ms or deemed significant enough to warrant discontinuing their infusion.

Twenty patients treated with the 2-h infusion and three patients treated with the 24-h infusion had plasma collected for pharmacokinetic studies. Tables 3 and 4 contain the summary of the pharmacokinetic results for patients treated with the 2-h infusion and the 24-h infusion, respectively. Patient nos. 19 and 20 received approximately half of their prescribed total dose of drug at 445 mg/m² before developing their life-threatening cardiac arrhythmia. Thus, their pharmacokinetic results do not show the proportional increase in peak drug and AUC_0-t levels expected had they received their full dose of drug. As expected, the peak drug concentrations were approximately 3–10-fold less with the 24-h infusion compared with the 2-h infusion, but the AUC_0-t levels have similar ranges. The terminal t_1/2 range was 1.4–22.5 h, but for most it was ≤5 h. A comparison of the SR233377 plasma concentrations obtained in patients (one on each schedule)treated with 225 mg/m² over 2 h (no. 13) or 24 h (no. 22) is shown in Fig. 2.

Among the 18 patients who completed their 2-h infusion, the statistical relationship of peak plasma concentration to total dose and then AUC_0-t to total dose was explored. The quadratic and cubic terms in total dose were not significantly related to either peak plasma drug concentration or to log₁₀ AUC _0-t; hence, their best model was a linear one. As shown in Fig. 3, total dose had a highly statistically significant linear regression relationship (P < 0.0001) to peak plasma drug concentration. The dashed lines indicate the 90% prediction limits for an individual peak plasma drug concentration level at a given total dose. For example, at a SR233377 total dose of 450 mg (an actual total dose used in the study), the model predicted a peak plasma drug concentration of 1622 ng/ml. The 90% prediction interval for the peak plasma drug concentration of an individual patient is from 959 to 2285 ng/ml. Thus, for a patient given 450 mg total dose, their observed peak plasma drug concentration would fall within these limits 90% of the time. This model had a R² = 77%,indicating that total dose could explain 77% of the variability in peak plasma concentrations.

Log₁₀ AUC _0-t was also best fit by a linear regression model (Fig. 4) in which total dose was highly statistically significant(P < 0.0001). This model also had an R² value of 77%. At a SR233377 total dose of 450 mg, the model predicted an AUC_0-t of 7,811 ng/ml × h. The 90% prediction interval for the AUC_0-t of an individual patient treated at 450 mg total dose is from 3,085 to 19,779 ng/ml × h.

We also performed a regression modeling analysis using dose level(mg/m²) instead of total dose (mg). The results were very similar: a highly significant linear effect(P > 0.0001) of dose and no significant quadratic or cubic effects, on either peak plasma concentration of drug or on(log₁₀ of) AUC. Model fit was also similar to the models using total dose: R² = 78% for peak plasma concentration of drug and R² = 75% for(log₁₀ of) AUC. At a dose level of 225 mg/m² (one actually used in our dose escalation),these regression models predicted a mean peak plasma drug concentration of 1,564 ng/ml and an AUC of 7,079 ng/ml × h.

SR233377 is a novel cytotoxic agent, the mechanism of action of which appears to be related to topoisomerase II inhibition. Preclinically, in mice, the acute toxicity was seizure-like activity with rapid i.v. administration, which was ameliorated with a short-term(10-min) infusion. Given the cardiac toxicity observed in the clinical trial and other preclinical models, cardiac arrhythmia may have been responsible for the seizure-like activity noted in mice. However, with this schedule alteration, the dose-limiting toxicity in mice was myelosuppression. Preclinical toxicology confirmed this as the dose-limiting toxicity using an in vitro myelotoxicity assay (5). This assay demonstrated greater sensitivity of mouse marrow (IC₇₀, 500 nm)compared with human marrow (IC₇₀, 1300 nm) with a 1-h exposure.

Because cardiotoxicity was identified during the initial portion of the clinical trial, significant work was done in preclinical models to determine the optimal model for evaluation of the cardiotoxicity. This optimal model could then be used to better define the cardiotoxic effects of SR233377 and as an additional screening tool for analogue search and development to find an agent devoid of cardiac effects. Preclinical electrophysiological studies using the rabbit ventricle model demonstrated that SR233377 caused lengthening of the action potential duration in both conductive and contractive tissues. This lengthening of the action potential duration of both the Purkinje and ventricular fibers strongly suggested a decrease in cellular membrane potassium permeability during depolarization and repolarization phases. This effect occurred at 10 μm and was concentration dependent. The resting potential, maximum velocity of initial phase of the action potential, and action potential amplitude were not modified by SR233377 (10).

The phenomena of prolongation of QT_C, followed by torsades de pointes, has been observed with antineoplastic agents as well as other classes of compounds. Amsacrine, an antineoplastic agent with structural similarities to SR233377(-NH-SO₂CH₃ side chain),has been reported to cause similar cardiotoxicity (11). Acodozole was another antineoplastic agent whose clinical development was aborted in Phase I because of torsades de pointes and QT_C prolongation (12, 13). Acodozole, however, had no structural similarities to SR233377. Several of the other non-antineoplastic agents reported to cause torsades de pointes (e.g., nonsedating histamine blockers, erythromycin,and ketoconazole) also do not have structural similarities to SR233377 (14). Typically, when torsades de pointes and QT_C abnormalities are drug induced, the cardiac rhythm will revert to normal spontaneously when the causative agent is discontinued, as was the case with the 2 patients who developed arrhythmias treated in our trial at the 445 mg/m²dose level.

Interpatient variability of pharmacokinetic parameters was observed,but generally the mean peak plasma concentration and AUC_0-t of SR233377 increased with dose. Significant linear statistical relationships of total dose to peak plasma level and to log₁₀ AUC _0-t were found, although derived from only 18 patients completing the 2-h infusion. The metabolism of[¹⁴C]-SR233377 has been investigated in vitro using a panel of 10 human liver microsome samples to determine the major CYP P-450 isoenzymes involved. SR233377 was metabolized to one major and two minor metabolites. The production of the major metabolite correlated well with the activity of CYP 2D6. Studies coincubating SR233377 with quinidine (a known substrate of this enzyme) resulted in almost complete inhibition (>95%) of SR233377 metabolism, thus confirming CYP 2D6 involvement (15). These studies demonstrate that SR233377 is metabolized by a polymorphically expressed CYP P-450 and therefore may be subjected to either extensive or poor metabolic clearance in patients depending on individual isoenzyme expression. The activity of CYP 2D6 in patients was not investigated in this study.

Another Phase I study at Fox Chase Cancer Center was conducted using a daily 2-h infusion for 5 consecutive days (daily × 5) repeated every 35 days (16). Twenty-five patients were treated with the dose range of 4.8–74.7 mg/m²/day. The toxicities reported from this trial were neutropenia, fever, nausea,and dyspnea, with none of these being dose limiting. After significant cardiac toxicity was noted in our trial, cardiac monitoring was begun in the daily × 5 trial at the 74.7-mg/m²/day dose level. Asymptomatic prolongation of the QT_C was noted and prompted closure of the trial. Peak drug concentrations in plasma increased linearly with dose but showed substantial interpatient variability in AUC, clearance, and half-life. The increases observed in the peak drug concentrations and AUC_0-t of SR233377 showed no evidence of drug accumulation over the 5 treatment days based on the comparison of day 1 and 5 results of these pharmacokinetic parameters. Tumor growth inhibition lasting for 4 months or more was observed in 7 patients.

The cardiotoxicity of SR233377 has terminated further development of the drug. Despite a schedule alteration to 24 h, which was implemented to minimize the peak plasma drug levels,QT_C prolongation still resulted. Thus, this toxicity does not appear to be related to peak drug concentrations in plasma. The search for an analogue has been initiated, and several analogues have similar preclinical in vitro and in vivo antitumor activity (1). In vitrotoxicity assays for myelotoxicity and cardiotoxicity will be used in hopes of identifying an analogue without cardiotoxicity prior to its Phase I clinical investigation. It appears from preclinical investigation that two possible candidates for analogue development have been identified. We await additional testing prior to the determination of which candidate will enter Phase I clinical trials.