A Phase I and Pharmacologic Evaluation of the DNA Intercalator CI-958 in Patients with Advanced Solid Tumors¹

DNA intercalating agents, including anthracyclines such as doxorubicin, have been used for many years in the treatment of patients with cancer. However, prolonged treatment with many of these compounds can result in cardiotoxicity (1). The search for agents that provide therapeutic benefits with less toxicity led to the development of a new class of intercalating agents, the BTPIs.³ The BTPI agents differ structurally from the anthrapyrazoles in that they contain a sulfur in place of the carbonyl in the center ring. This modification may lessen the cardiotoxic potential of this class of compounds by reducing the possibility for the semiquinone free radical generation thought to be involved in anthracycline-induced cardiotoxicity.

The most active member of the BTPI class, with respect to both degree and spectrum of anticancer activity, is CI-958 (Fig. 1). CI-958 is a stable, synthetic, highly water-soluble drug. Preclinical data for CI-958 demonstrated marked antitumor activity against a broad range of murine and human tumors including leukemia, melanoma, lung, colon, and breast cancer (murine tumor models included P388, L1210, B16, M5076, as well as mammary 16c,17, 13c, 25, colon 36, 11a, and Ridgeway and Glasgow osteosarcoma;human tumor xenografts included MX-1 mammary carcinoma and LOX melanoma). In general, its activity resembles that of doxorubicin with superiority in some systems including one mammary and two colon murine tumor models. When compared with mitoxantrone, amsacrine, and the anthrapyrazoles, CI-958 has a broader spectrum and higher degree of activity against a panel of murine carcinoma and human tumor xenografts.⁴ In addition, in vitro studies with multidrug-resistant tumor lines have shown that CI-958 is less cross-resistant than other intercalating agents such as doxorubicin (2). Furthermore,development of acquired resistance to CI-958 is rare in human breast cancer cell lines that rapidly develop resistance to doxorubicin and mitoxantrone.⁴

Similar to doxorubicin and mitoxantrone, CI-958 is a potent inhibitor of nucleic acid synthesis, suppressing RNA and DNA synthesis to an equal extent. Tight DNA binding via intercalation between the base pairs is the proposed mechanism of action. This binding is thought to lead ultimately to breakage of the DNA, resulting in cell death. Like doxorubicin, CI-958 produces both single- and double-strand protein-associated DNA breaks that undergo repair very slowly. However, in vitro studies show that unlike doxorubicin and most other quinone or quinone-derived antitumor agents, CI-958 demonstrates little potential to generate superoxide radicals (3). Preclinical toxicology studies showed that CI-958 produced fewer cardiac lesions in rats than did doxorubicin. The dose-limiting toxicity of CI-958 in rats was myelosuppression.⁴ Because of its broad spectrum and high degree of activity as well as a favorable toxicity profile, CI-958 was chosen for further development. The Phase I study described here was undertaken to determine the toxicity profile,maximum tolerated dose, and pharmacokinetics of CI-958 given as an i.v. infusion every 21 days in patients with refractory solid tumors.

Over the past decade, there has been increasing interest in the development of new Phase I trial designs that minimize the number of patients receiving biologically inactive doses of the Phase I agents without greatly increasing the risks of toxicity. For example, Simon et al. (4) used mathematical models based on data from 20 Phase I trials to evaluate four novel dose escalation designs. Their analysis suggests that accelerated titration designs using rapid interpatient dose escalation will effectively reduce the number of patients who are undertreated and speed the completion of Phase I trials without significantly compromising safety (4). O’Quigley et al. (5) proposed the continual reassessment method, which uses clinicians’ estimates of the dose range which may be toxic. Dose escalation is rapid with toxicity evaluations of each cohort determining the dose of the next cohort (5). The PGDE design used in the present study is an example of a novel trial design that has been proposed to lessen the number of patients treated at inactive dose levels and shorten the time required to complete the Phase I trial (6). This dose escalation design was first proposed in 1986; it is based on the concept that interspecies differences in drug metabolism, elimination,and binding are largely responsible for interspecies differences in toxicity. Therefore, the AUC at the mouse LD₁₀may better approximate the AUC at human MTD than the mouse LD₁₀ approximates the human MTD. The PGDE design uses real-time pharmacokinetic analysis and comparison with preclinical models and targets dose escalation to reach an AUC equivalent to that seen at the murine LD₁₀ within three to four steps. Prior attempts to use PGDE have been hampered by insensitive assays, interspecies differences in metabolism or target cell sensitivity, and interpatient variability in clearance.

Patients with refractory solid tumors were recruited from the Johns Hopkins Oncology Center Outpatient Department. Eligibility criteria included: (a) life expectancy of at least 12 weeks;(b) a Zubrod performance status score of two or better (7); (c) age 18 years or older and not pregnant or breast feeding; (d) ability to give informed consent;(e) no more than one prior regimen containing an anthracycline-like compound and prior cumulative doses no greater than 300 mg/m² of doxorubicin or 125 mg/m² of mitoxantrone; (f) adequate renal (serum creatinine less than 2 × upper limit of normal or 24 h creatinine clearance >60 ml/min), hepatic (serum bilirubin less than 1.5 × upper limit of normal), and hematological function (granulocyte count >1500/mm³ and platelet count >100,000/mm³); and (g)no history of myocardial infarction, angina, cardiomyopathy, or ventricular arrhythmia. Additionally, patients must not have had chemotherapy or radiation therapy within the 3 weeks prior to enrollment and must have fully recovered from the toxic effects of previous regimens. The clinical trial described was carried out with approval from the institutional review board. Written informed consent was obtained from all patients prior to study entry according to institutional and federal guidelines.

In this dose escalation study, CI-958 was administered as an i.v. infusion every 3 weeks. The starting dose was 5.2 mg/m², which corresponds to one-tenth of the mouse LD₁₀ from preclinical single-dose toxicity studies.⁵ At least 3 patients were treated and evaluated at each dose level. Escalation to the next dose level was permitted only after 2 of the 3 patients treated at each level had been monitored for a minimum of 3 weeks and the third patient a minimum of 2 weeks. Results of the pharmacokinetics from all 3 patients were obtained prior to proceeding to the next dose level.

The dose escalation scheme used in this trial was based on the concept proposed by Collins et al. (6). This PGDE scheme targets an AUC equal to the murine AUC at the LD₁₀ and proceeds in three stages (Table 1). In stage 1, the median CI-958 AUC in patients was determined at the starting dose. The first escalation step used the geometric mean method. By this method, the calculated dose for the second dose level was equal to the starting dose multiplied by the square root of the ratio of the AUC in mice at the LD₁₀ to the median AUC in humans at the entry dose level (obtained from the three patients at the first dose level). If the calculated dose for level 2 represented a >3-fold increase above level 1, dose level 2 was limited to a 3-fold increase. In stage 2, if the median CI-958 AUC in humans at dose level 2 was <40% of the murine AUC at the LD₁₀, then dose levels 3 and higher were determined by the extended-factors-of-two method and were each equal to a 100% increase above the preceding dose level until either: (a) the median CI-958 AUC in patients had reached 40% of the murine AUC at the LD₁₀; or(b) two of the three patients treated at a dose level experienced grade 2 drug-attributable adverse events. In stage 3 of this dose escalation design, if the target 40% AUC was reached without grade 2 toxicity, then dose escalation continued using the modified Fibonacci schema. If grade 2 toxicity was seen, then subsequent doses were only 1.33 times the previous dose level.

CI-958 was supplied as a lyophilized powder in 50- or 250-mg vials. Vials containing 50 mg of drug were reconstituted with 5 ml of sodium phosphate buffer and diluted in D5W. The CI-958 contained in the 250-mg vials was formulated to contain a buffer system, eliminating the need for a separate buffer solution, and was reconstituted in sterile water for injection. Upon reconstitution, doses of CI-958 were initially diluted into 100 ml of D5W and administered over a period of 1 h. When local venous irritation became a problem, doses were diluted to a total volume of 400 ml in D5W and infused over a period of 2 h.

Retreatment occurred every 21 days provided the patient had not experienced dose-limiting toxicity and had fully recovered from the previous course. Dose escalations were permitted in individual patients at the completion of at least two 3-week courses if the prior course of treatment did not result in any unacceptable toxicity or tumor progression, if all eligibility criteria continued to be met, and if patients previously untreated with CI-958 had already been evaluated at the higher dose. Dose reductions for subsequent courses in individual patients were based on toxicity. If a patient experienced grade 4 hematological or grade 3–4 nonhematological toxicity, the next dose was reduced to the preceding dose level. Patients who had not recovered by day 21 had subsequent therapy delayed weekly until recovery.

Complete history, physical examination, assessment of performance status, routine laboratory studies, electrocardiogram, and measurement of sentinel tumor lesions were conducted for each patient within 7 days prior to first treatment. While on study, patients were followed weekly with complete blood count and differential, serum electrolytes, and serum chemistry profile (bilirubin, aspartate aminotransferase, alanine aminotransferase, alkaline phosphatase, BUN, and creatinine). A complete physical examination was repeated prior to each course of treatment. Toxicity was evaluated for each dose level and each course of therapy and was monitored on an ongoing basis. Appropriate radiographic and laboratory studies were performed as necessary to follow the disease response to treatment. Measurements of sentinel lesions were reported every 3 weeks for lesions detected by palpation or chest X-ray or every 9 weeks for lesions, followed by CT scans.

In this study, dose-limiting toxicity was defined as a granulocyte nadir <500/mm³, a platelet count nadir<50,000/mm³, grade 3 or 4 nonhematological toxicity, or grade 2 neurological, renal, or cardiac toxicity. The MTD was defined as that dose level of CI-958 that produced dose-limiting toxicity in at least two patients treated at that dose level. When the MTD had been determined, additional patients were treated at the level below the MTD to better define the recommended Phase II dose.

Venous blood samples (5 ml) were collected prior to and 0.25, 0.5, 1(end of infusion), 1.25, 1.5, 1.75, 2, 2.5, 4, 6, 8, 12, 24, 32, 48,72, 96, 120, 144, 168, 336, and 504 h after the start of the infusion. When infusion length was changed from 1 to 2 h, the 0.25, 0.5, 1.25, 1.5, and 1.75 time points were omitted. To anticoagulate and enhance plasma stability of CI-958, samples were immediately transferred to plastic centrifuge tubes to which 0.45 m citrate buffer (pH 5.0) containing 40 mg/ml ascorbic acid had been added (25 μl/ml blood). Samples were gently mixed and centrifuged as soon as possible, and plasma obtained was rapidly frozen. Because CI-958 undergoes photodecomposition, plasma samples were kept from light. Urine excreted in the first 24 h after CI-958 administration was collected in opaque bottles in three intervals of 8 h each. Thereafter, urine was collected in 24-h intervals for the subsequent 168 h.

Both RIA and HPLC methods were initially used to measure the concentration of CI-958 in plasma and urine samples. Plasma CI-958 concentration-time data obtained from the RIA method was higher than that determined using HPLC, suggesting that the CI-958 and its metabolites were quantified simultaneously by RIA. Therefore, the pharmacokinetic results reported are those obtained using the HPLC data. The pharmacodynamic relationship between AUC and toxicity correlated equally well when AUC was determined by RIA or HPLC,suggesting that no information was lost by using HPLC. Plasma and urine samples were assayed for CI-958 concentration at BAS Analytics (West Lafayette, IN) according to a validated HPLC method using electrochemical detection.⁶

Briefly, the method included solid-phase extraction of plasma samples on Bond-Elut C₂ cartridges from which CI-958 was eluted with 0.1% ethylenediamine eluting solvent. The eluent was dried, and the residue was reconstituted in 0.5–4.0 ml of the mobile phase, depending on the estimated plasma concentration. A 100-μl aliquot was injected on the HPLC system. Chromatographic separation was performed on a DuPont Zorbax Rx 4.6 × 250 mm C-8 column at 40°C. The mobile phase consisted of a mixture of methanol/n-propyl alcohol (2:1 for plasma analysis and 1:1 for urine analysis) and buffer (0.07 m citrate buffer for plasma analysis and 0.7 m citrate buffer for urine analysis). The mobile phase was run isocratically at 1.5 ml/min. Peaks of interest were detected electrochemically at an applied potential of +525 mV (plasma) and +400 mV (urine). CI-958 concentrations were quantitated by the peak height ratio method using PD-112451 as the internal standard. System reproducibility expressed as relative SD (%) of the peak height ratios was determined using pooled human plasma extracts. The reproducibility of the HPLC system (%RSD)was 1.6, 0.7, and 1.7% for CI-958 concentrations of 5.0, 100, and 1000 ng/ml. Assay precision and accuracy were determined by analyzing three quality control pools in triplicate over three separate days. Assay precision expressed as relative SD (%) of the assayed concentrations was 4.7, 6.2, and 3.6% for quality controls containing 5, 500, and 5000 ng/ml CI-958, respectively.

CI-958 pharmacokinetic parameter values were calculated after each dose administered using noncompartmental methods. Length of infusion times and samples collection times varied in each patient, and hence, actual collection and infusion times were used in the analysis. AUC was determined by Lagrange polynomial interpolation (8). The apparent elimination-rate constant was estimated as the absolute value of the slope of a linear regression of natural logarithm (ln) of plasma CI-958 concentration against time during the terminal phase of the plasma concentration-time profile. The terminal elimination phase was determined by visual inspection. Apparent elimination half-life was calculated as ln (2) divided by the elimination rate constant. Total plasma clearance was calculated as dose/AUC_(0-∞). Volume of distribution at steady state was calculated as [Dose ×AUC_(0-∞)/AUC_(0-∞)²− ((Dose × T)/2 ×AUC_(0-∞))], where T is the duration of the constant rate i.v. infusion. To determine whether a relationship exists between CI-958 exposure and toxicity, the percent reduction in ANC was plotted versusAUC, and the data were fit with a simple Emax model (9)using PCNONLIN 4.0 (SCI Software, Lexington, KY) and WinNonlin 1.1(Scientific Consulting, Inc., Apex, NC).

Forty-four patients received 162 courses of CI-958. The patient characteristics and demographics are displayed in Table 2. Eleven dose levels were evaluated (as listed in Table 3). Dose level 1 was set at 5.2 mg/m², corresponding to one-tenth of the murine LD₁₀. Dose level 2 was limited to 3-fold dose level 1 because the dose determined by the geometric mean calculation (17 mg/m²) exceeded the defined limit(15.6 mg/m²). Dose levels 3 and 4 were reached by doubling the preceding dose level, because the AUC values obtained at dose levels 2 and 3 did not reach the target AUC value of 3192 ng·h/ml (40% of the murine AUC at the LD₁₀). At dose level 4, the target AUC was reached, using both HPLC and RIA methodology. Therefore, dose levels 5–9 were escalations based on the modified Fibonacci scheme. The last two dose levels used smaller increments based on the degree of neutropenia observed at level 9.

The number of courses of treatment per patient ranged from 1 to 29. Thirty-six patients (82%) received more than one course, and 8 patients (18%) received five or more courses. Four patients were escalated to higher dose levels for subsequent courses, and 4 patients were dose reduced. All patients were evaluable for toxicity. Patients were withdrawn from the study because of progressive disease (38 patients; 86%), death within 30 days of last dose (3 patients; 7%;all described in detail below), occurrence of adverse events (2 patients; 5%), or refusal of further treatment (1 patient).

Severity of neutropenia was dose related. Median ANC nadir and maximum toxicity grade per course are listed by dose level in Table 3. Neutropenia was dose limiting (grade 4) in 1 of 7 patients at 425 mg/m² (1 of 28 courses), 2 of 8 patients (3 of 18 courses) treated at 560 mg/m², 4 of 9 patients (6 of 21 courses) treated at 700 mg/m², and 2 of 6 patients treated at 875 mg/m². Neutrophil count nadirs occurred at a median of 14 days (range, 8–23 days), and the median day of recovery (ANC ≥1500/mm³) was 22 days (range, 12–29) after dosing. One of the two patients mentioned above with dose-limiting neutropenia at 560 mg/m²had also had dose-limiting neutropenia at 700 mg/m² and had required dose reduction. Similarly,one of the patients with dose-limiting neutropenia at 700 mg/m² had been treated previously at 875 mg/m² and had required dose reduction for dose-limiting neutropenia.

Two patients developed serious adverse events related to their neutropenia. One patient treated with 700 mg/m²CI-958 developed neutropenia, fever, and sepsis. She was hospitalized and begun on antibiotics. Two days later, the antibiotics were discontinued according to the wishes of the patient and her family. The patient died 2 days later. Another patient required hospitalization for febrile neutropenia after her second course of CI-958 (875 mg/m²). She was treated with broad spectrum i.v. antibiotics. All cultures were negative; no infection was detected. Her neutropenia resolved, and she was discharged. Both of these serious adverse effects were considered related to CI-958.

In general, thrombocytopenia was infrequent and mild. In total, 6 patients experienced thrombocytopenia of any grade. Thrombocytopenia appeared to be dose related. Median platelet nadir and maximum toxicity grade per course are listed by dose level in Table 3. Two patients treated at 700 mg/m² developed grade 3 or 4 thrombocytopenia. One patient developed grade 3 and grade 4 thrombocytopenia after each of 2 courses of CI-958 at 875 mg/m². Thus, thrombocytopenia was dose limiting in 1 of 8 patients at 700 mg/m²and 1 of 6 patients at 875 mg/m².

Treatment-associated anemia was infrequent. Nine patients (20%)experienced anemia of any grade. Three of these were considered to have treatment-associated anemia. One patient at 700 mg/m² and 1 patient at 875 mg/m² had grade 1 anemia. One patient treated at 700 mg/m² had grade 3 anemia.

The most common treatment-associated, nonhematological toxicities were nausea and/or vomiting, asthenia, fever, chills, diarrhea, injection site symptoms (inflammation, reaction, edema, or pain), phlebitis, and vasodilation. Table 4 shows the distribution of these toxicities by dose level and toxicity grade. Dose levels 1–5 (5.2–125 mg/m²) are not depicted because no toxicity greater than grade 1 was observed at these dose levels. As shown, most of the toxicities seen at doses 200 mg/m² and above were mild or moderate. One patient treated with 560 mg/m² had grade 3 injection site symptoms. One patient treated at 700 mg/m² had grade 3 fever on the day of infusion. Two patients had grade 4 nausea and vomiting at doses of 560 and 875 mg/m².

Infusion of CI-958 through peripheral veins caused local reactions characterized by phlebitis, induration, erythema, or pain. These problems were temporarily alleviated by increasing the volume of D5W infused with each dose and lengthening the infusion duration from 1 to 2 h. However, as the dose levels increased, local venous intolerance was again encountered. Use of central venous catheters successfully prevented these problems.

One case of CI-958 extravasation occurred. The patient did not complain of local discomfort during the infusion but thereafter developed ulceration at the site of the injection. This reaction developed slowly over a period of several weeks, eventually resulting in several small areas of necrosis and was very slow to heal. A subsequent treatment in the opposite hand caused an exacerbation at the initial reaction site.

A variety of acute reactions were reported during infusions of CI-958,characterized by flushing, urticaria, pruritis, and hyperesthesia,particularly of the head and neck region. These reactions were dose related and appeared to be related to the rate of infusion. On several occasions, infusions were either slowed or temporarily interrupted, the reactions subsided, and the infusions were resumed to completion. Reactions occurred sporadically, sometimes during the initial treatment(7 patients) and sometimes occurring only after multiple courses were given (4 patients). Reactions generally were self-limited and often did not recur on subsequent exposure.

In addition to the common nonhematological toxicities discussed above,two patients experienced dose-limiting renal or hepatorenal toxicity associated with CI-958 treatment at 875 mg/m². One patient, who had mildly elevated creatinine secondary to prior cisplatin and etoposide chemotherapy, was noted to have acute renal failure on day 8 after the second course of CI-958 treatment at 875 mg/m². His baseline BUN and creatinine prior to course 1 were 26 and 1.9 mg/dl, respectively. Prior to course 2, his BUN and creatinine were 31 and 1.8 mg/dl, respectively. He developed severe nausea, vomiting, and decreased oral intake after his second treatment. BUN and creatinine rose to 130 and 14 mg/dl on day 10. The patient was hospitalized with acidosis and uremia and treated with hemodialysis. He subsequently required mechanical ventilation and pressor support. He died on day 12 without recovery of his renal function. Another patient treated with CI-958 at 875 mg/m² required hospitalization for acute renal insufficiency (48 mg/dl BUN and 2.5 mg/dl creatinine) and abnormal liver function tests (6.4 mg/dl peak bilirubin, >2400 IU/l transaminases) after her first course of therapy. She was treated with aggressive supportive care and recovered. No other patients experienced elevations in creatinine over 1.5 times normal (grade 1). Four other patients had transient mild to moderate transaminitis during treatment that resolved to baseline.

Patients were carefully monitored for any signs of cardiotoxicity. No clinically significant electrocardiogram changes were noted after treatment with CI-958 at any dose level. No patients developed signs or symptoms of cardiomyopathy after treatment with this agent. One patient with lung cancer experienced atrial fibrillation, pericardial effusion,and shortness of breath on day 1 after treatment with CI-958 at 875 mg/m². He was treated with digoxin and quinidine gluconate, and the atrial fibrillation resolved. The atrial fibrillation was thought possibly attributable to CI-958.

Pharmacokinetic sampling was performed on 38 patients. Individual plasma CI-958 concentration-time profiles exhibited a multiexponential decline at the end of infusion. Concentration-time curves were characterized by secondary peaks occurring throughout the distribution and elimination phases. A representative example, the concentration-time curve from patient 33 who received 700 mg/m², is shown in Fig. 2. Mean CI-958 pharmacokinetic parameters are summarized in Table 5. Plasma CI-958 AUC increased proportionally with dose (Fig. 3). However, there was considerable interindividual variability, which resulted in similar AUC values in individuals treated with widely differing doses of CI-958. At doses where biological activity was observed (dose ≥200 mg/m²), 2–11% of the AUC was contributed by the extrapolation from the last sample to ∞. The clearance values were similar at all doses, consistent with linear pharmacokinetics of CI-958(Fig. 4). Concentrations in the terminal elimination phase in many patients were near the limit of quantitation and exhibited secondary peaks. Therefore, the elimination half-life values reported may be underestimated and should be interpreted with caution. The cumulative amount of unchanged CI-958 excreted in urine of each individual was <10% of dose, indicating that urinary excretion of CI-958 is a minor elimination pathway.

Two patients experienced dose-limiting renal or renal/hepatic toxicity associated with CI-958 treatment as discussed above. Plasma CI-958 AUC values for these patients during the first course of therapy were twice those of other patients treated with 875 mg/m²dose of CI-958 (Fig. 3). Therefore, it would appear that these toxicities may have resulted from higher systemic exposure to CI-958. Plasma CI-958 AUC values also predicted the reduction in ANC. However,AUC was no more effective in predicting neutropenia than was dose. The relationships between percent reduction in ANC from baseline versus plasma CI-958 AUC values and CI-958 dose are depicted in Figs. 5 and 6.

There were no tumor responses after treatment with CI-958. However,patients were permitted to remain on treatment as long as their disease was stable. Eight patients received 5 or more courses of therapy. One patient with breast cancer remained on study for 29 courses (609 days),and one patient with colon cancer remained on study with stable pulmonary disease for 12 courses (252 days).

The dose-limiting toxicities of CI-958 were neutropenia and renal or hepatorenal toxicity. Grade 4 neutropenia was dose related (Fig. 6),with four episodes at doses below 700 mg/m² and eight episodes at doses of 700 mg/m² and above. The only two episodes of nonhematological dose-limiting toxicity occurred at 875 mg/m². The MTD of CI-958 was determined to be 875 mg/m², based on the renal and liver toxicity, as well as the dose-limiting myelosuppression encountered at this dose. The preceding dose level, 700 mg/m², was well-tolerated by patients who had not been heavily exposed to prior chemotherapy and is the recommended Phase II starting dose for those patients. In heavily pretreated patients,560 mg/m² is the recommended Phase II dose.

The BTPI class of compounds, and CI-958 in particular, was developed in part because of structural changes hypothesized to make these agents less cardiotoxic than anthracyclines. In this Phase I study, one patient treated at the MTD developed atrial fibrillation in the setting of advanced lung cancer. No other patients experienced any electrocardiogram abnormalities or other evidence of cardiotoxicity. Although preclinical models suggest that CI-958 may be less cardiotoxic than doxorubicin,⁴ accurate estimation of the cardiotoxicity of this compound can only be made after further testing.

CI-958 has shown marked antitumor activity in preclinical models. Although assessing efficacy was not the primary end point of the present study, no tumor responses were seen. However, eight patients received five or more courses, and two patients had prolonged periods of stable disease. Preliminary Phase II evaluation of CI-958 at the doses recommended above has been carried out. In a pilot Phase II study in patients with hormone refractory prostate cancer, CI-958 was given at a dose of 700 mg/m² over 2 h every 3 weeks. Six of 30 patients with elevated prostate-specific antigen had response ≥50% reduction from baseline. Two of 11 patients with measurable disease responded (10). Another Phase II study has evaluated CI-958 in 15 patients with colorectal cancer and did not find this dose and schedule to be effective (11).

PGDE has been proposed as a potentially safer and faster dose escalation method. Its utility in many situations was confirmed by retrospective analysis of results from a number of Phase I trials (6). However, there are a number of situations in which this dose escalation technique cannot be used effectively. For example,PGDE cannot be used effectively if there are interspecies differences in target cell sensitivity or schedule dependence or differences in metabolism or plasma protein binding that may not be accurately reflected by measuring total plasma levels of parent drug. Additionally, there are a number of technical factors that may limit the applicability of this technique, such as differences in the mode of drug delivery between preclinical and clinical testing, or limitations of the assay in accurately measuring concentrations at both the LD₁₀ and one-tenth LD₁₀doses (6, 12).

These caveats to the use of PGDE have been borne out by prior use of this technique in clinical trials. For example, a number of retrospective reviews of the experience with antimetabolites have shown that many of these compounds are not amenable to PGDE trial design because of profound interspecies differences in target cell enzyme properties (6, 13). By contrast, the anthracyclines appear to be a class of drugs for which the AUC is generally a much better predictor of toxicity than dose, and thus this class should be amenable to successful use of PGDE. However, the prospective use of PGDE in a Phase I trial of the new anthracycline 4′-iodo-4′-deoxydoxorubicin was limited by unexpected interspecies differences in the metabolism of 4′-iodo-4′-deoxydoxorubicin. Near the end of the trial, when the target for escalation was redefined as the sum of the concentration of both the parent drug and the active metabolite, PGDE could be used. However,the number of dose escalation steps was only one fewer than the modified Fibonacci plan because PGDE could only be used late in the trial (14).

Several prior Phase I trials of anthrapyrazoles have attempted PGDE with varying results. For example, wide interpatient variability in AUC made application of PGDE impossible in a Phase I trial of the anthrapyrazole CI-941 (15). In a Phase I trial of the anthrapyrazole, piroxantrone, administered as a 1-h infusion every 3 weeks to patients with advanced cancer at our institution (16), the usefulness of PGDE was limited by a relatively insensitive assay, rapid plasma clearance not anticipated by the sampling scheme, and close proximity of the AUC at the starting dose to the target AUC, where Fibonacci escalation should begin. Because of these limitations, there was no reduction in number of dose escalation steps from that required with a Fibonacci approach. In a similar Phase I trial of piroxantrone conducted at a different institution,utilization of PGDE was also hampered by assay insensitivity and interspecies differences in clearance such that the first dose escalation steps had to be made empirically. Despite these problems,this study required 6–9 fewer patients than would have been needed to reach the MTD using the Fibonacci approach (17).

Interspecies differences in pharmacokinetics may also have influenced the accuracy of PGDE in the present study. The AUC at the human MTD was 6.6 times higher than the AUC at the murine LD₁₀, not identical as hypothesized. Thus, even when compared by AUC instead of dose, the mouse overpredicted human toxicity. The cause of this disparity is unclear but may be explained in part by interspecies differences in plasma protein binding or pharmacokinetics such as higher clearance rates and larger volume of distribution in humans compared with mice (18). Nonetheless, this study represents the most successful use of PGDE to date.

The PGDE design used in the present study proved to be an effective and efficient method for reaching the MTD of CI-958. The MTD obtained in patients (875 mg/m²) was 168-fold greater than the starting dose (5.2 mg/m²), which had been calculated based on preclinical toxicology data. Only 11 dose levels were needed to reach the MTD by using the PGDE design. Using the modified Fibonacci method, dose level 11 would be only 38 times greater than the starting dose, and an additional 5–6 dose levels (15–18 patients) would have been required to reach the MTD of 875 mg/m². Therefore, using the PGDE approach substantially decreased the total number of patients studied and the length of time required for this study. Of note, the 3-fold escalation for the first step in this study is, to our knowledge, the largest increment reported in any Phase I cancer chemotherapy trial, regardless of design. Importantly, because of the rapid early escalation, the savings in patient resources occurred at early dose levels that were the least likely to be effective. This may represent the most successful use of PGDE to date. However, the future application of this method remains in doubt as a result of the overall experience to date and the emergence of other accelerated titration designs.