Celecoxib and Mucosal Protection: Translation from an Animal Model to a Phase I Clinical Trial of Celecoxib, Irinotecan, and 5-Fluorouracil

Irinotecan in combination with 5-fluorouracil (5-FU) and leucovorin is a commonly used regimen for metastatic colorectal cancer (1). The ‘IFL’ regimen, consisting of bolus 5-FU, leucovorin, and irinotecan, was associated with an overall response rate of 35% to 39 % and an improvement in overall survival when compared with 5-FU and leucovorin. This regimen was limited by grade III/IV diarrhea, neutropenia, mucositis, and early mortality (2). Infusional 5-FU-based combination regimens are popular in Europe and have similar efficacy but lower toxicity than the IFL regimen. The infusional regimen, FOLFIRI, consists of irinotecan: 180 mg/m² as a 90-minute infusion on day 1; leucovorin: 400 mg/m² as a 2-hour infusion during irinotecan infusion; and followed by 5-FU bolus: 400 mg/m² and a 46-hour infusion of 5-FU, 2.4 to 3 g/m² every 2 weeks (3). Although the infusional regimens have an improved toxicity profile compared with IFL, serious toxicities can occur with this therapy. Development of mechanism-based approaches aimed at selective amelioration of treatment-induced toxicity and optimization of dose is therefore of clinical importance.

Irinotecan-induced diarrhea is related to its metabolism by carboxylesterase to SN-38 (4). We hypothesize that SN-38-induced diarrhea through intestinal mucosal damage is associated with prostaglandin-mediated inflammation and that cyclooxygenase-2 (COX-2) is involved in the process. In pathologic conditions associated with increased COX-2 expression, the resulting additional prostaglandin E₂, in addition to stimulating mucus secretion, may also act on the epithelial cell lining of the mucosa, triggering chloride (Cl⁻) secretion, water loss, and diarrhea.

Trifan et al. (5) investigated the role of COX-2 in the pathogenesis of irinotecan-induced late diarrhea using a rat model. They reported that COX-2 is induced in the rat colon after irinotecan treatment and that this is concurrent with an increase in prostaglandin E₂ production. Histopathologic analysis of the colonic epithelium in their rat model revealed glandular cell dysplasia, mucosal atrophy, and fused villi. Results from their study indicated that celecoxib enhances the antitumor effect of irinotecan and reduces the severity of late diarrhea in a dose-dependent manner. We have extended these observations to the Ward colorectal cancer rat model and examined irinotecan administration on a daily × 3 and weekly schedule with celecoxib. As we show here, the toxicity was ameliorated and the antitumor activity was increased by virtue of allowing the animals to survive an otherwise lethal dose.

Based on these data, we designed a phase I study consisting of escalating doses of irinotecan and fixed doses of 5-FU and leucovorin in combination with celecoxib. We hypothesized that (a) celecoxib would ameliorate irinotecan-related diarrhea, (b) addition of celecoxib would permit safe dose escalation of irinotecan beyond the usual maximum tolerated dose (MTD), and (c) amelioration of diarrhea by celecoxib would not be secondary to alteration of pharmacokinetics of irinotecan or its active metabolites SN-38 and SN-38G.

Ward colorectal cancer model. The Ward colorectal carcinoma rat model was used for antitumor activity evaluation. Nonnecrotic tumor pieces (100 mg) were transplanted s.c. to Fisher rats (Fisher F344/N, female, 150-180 g), via trocar under anesthesia. Treatment was initiated 14 days later or when tumor sizes (weight) reached 3,000 to 3,500 mg. Irinotecan and celecoxib were provided by Pfizer, Inc. (former Pharmacia and Upjohn, Kalamazoo, MI).

Drug administration. Irinotecan was administered i.v., whereas celecoxib was given orally. Irinotecan was given by i.v. push with two schedules: (a) once daily for 3 days (daily × 3) and (b) once weekly for 4 weeks (weekly × 4).

Irinotecan doses were as follows: 100, 150, and 200 mg/kg/d for the daily × 3 schedule and for the weekly schedule. Celecoxib was given by the oral route (30 mg/kg/d) in two divided doses for 7 days with the daily × 3 schedule of irinotecan or for 22 days with the weekly schedule of irinotecan. The first dose was given 24 hours before irinotecan administration. These doses and schedule of celecoxib were established by Trifan et al. (5) as optimal. The MTD was defined as the highest drug dose that was associated with weight loss <20% and reversible toxicities and did not lead to lethality. Drug-induced toxicities (body weight loss, diarrhea, and lethality) were observed daily for a minimum of 3 weeks after drug treatment, thereafter, two to three times weekly.

Tumor assessment. Two axes (mm) of tumor (L, longest axis; W, shortest axis) were measured with the aid of a Vernier caliper. Tumor weight (mg) was estimated by the formula: tumor weight = 1/2 (L × W²). Tumor measurements were taken daily for the first 10 days and at least thrice weekly for the first 4 weeks after therapy and twice weekly thereafter. Complete tumor regression (CR) was defined as the inability to detect tumor by palpation at the initial site of tumor appearance for >2 months after therapy. Partial tumor regression (PR) was defined as ≥50% reduction in initial tumor size. Four rats per treatment group were included and experiments were repeated at least once.

Diarrhea assessment. Diarrhea in animals was graded as follows: grade 1, small amount of watery diarrhea that lasted for 3 to 5 days and was followed by full recovery, without lethality; grade 2, moderate to severe diarrhea that lasted >5 days without blood or mucus and could be followed by recovery; grade 3, severe diarrhea with mucus, without blood, no recovery occurred, and death followed; and grade 4, severe diarrhea with blood and mucus, no recovery occurred, and death followed.

Statistical evaluation. Statistical calculations were done with SAS version 9.1 (SAS Institute, Cary, NC). Survival was computed using Kaplan-Meier estimation. Statistical comparison of the survival distribution between two treatment groups was done with the log-rank test. The presence of synergy was tested as an interaction in an exact logistic regression model for treatment response; each rat was classified as a binary outcome for antitumor activity (CR or PR was a response; otherwise, no response). The four treatment groups were based on irinotecan level (100 or 200 mg/kg) and celecoxib administration (yes or no). Interaction was detected if one of the treatment groups contributed to the response rate beyond the main effect due to irinotecan or celecoxib in the regression model.

Eligible patients had metastatic cancer, Eastern Cooperative Oncology Group performance status of 0 or 1, life expectancy >6 months, adequate hematologic variables (hemoglobin, >9 g/dL; absolute neutrophil count, >1,500; WBC, >4,000; and platelets, >100,000/mm³), adequate biochemical variables (total bilirubin and creatinine within institutional limits and aspartate aminotransferase <2.5 times institutional limit), and age >18 years. Prior chemotherapy (one prior regimen) and limited radiotherapy (<3,000 cGy to marrow) were permissible. All patients were required to sign an informed consent document.

Patients with known hypersensitivity to sulfonamides, uncontrolled brain metastases, uncontrolled intercurrent illness, including ongoing infection, symptomatic cardiac disease, active gastric ulcers, gastrointestinal bleeding, uncontrolled inflammatory bowel disease, history of second malignancy (except for curatively treated carcinoma of the cervix in situ or nonmelanomatous skin cancer), pregnancy, lactation, or Gilbert syndrome were excluded from participation.

Clinical study design. This was an open-label, single-center, nonrandomized, dose-escalating, phase I study. All laboratory tests required to assess eligibility had to be completed within 7 days before the start of the treatment. Irinotecan, 5-FU, and leucovorin were administered every 2 weeks. Celecoxib was given twice daily throughout the study. However, during cycle 1, celecoxib was started on day 2 to allow day 1 irinotecan pharmacokinetics to serve as a control for pharmacokinetic studies on day 14 to explore any modulation of irinotecan pharmacokinetics by celecoxib. Only irinotecan was escalated in this study. There was no intrapatient dose escalation.

Irinotecan was administered as a 90-minute infusion, diluted in 250 mL D5W or 0.9% saline through peripheral or central access. At the same time, 400 mg/m² leucovorin was administered, diluted in 250 mL saline (0.9%) over 2 hours through peripheral or central access. This was followed by 400 mg/m² 5-FU bolus followed by 2.4 gm/m² 5-FU as a continuous infusion over 46 hours through a central access. The cycle was repeated every 2 weeks. Celecoxib (400 mg) orally administered bid was continued throughout the study period, starting on day 2. Premedications included 8 mg ondansetron and 10 mg dexamethasone; both administered i.v. Atropine (0.4 mg) was administered s.c. for patients who developed diaphoresis and bradycardia during irinotecan infusion or for early-onset diarrhea (6). Irinotecan (Camptosar, Pfizer, New York, NY) was commercially available as a 2-mL single-dose vial and as a 5-mL single-dose vial (20 mg/mL active ingredient; Pfizer). The drug was prepared for administration according to directions in the package labeling. 5-FU injection was commercially available (Roche Laboratories, Nutley, NJ) and supplied as 50 mg/mL, 10 mL vials. Celecoxib was supplied as 200 mg capsules by Pfizer Pharmaceuticals (manufacturer Searle, Chicago, IL).

Dose escalation and dose-limiting toxicities. Four dose levels (DL) were to be explored: 180, 200, 220, and 260 mg/m². Dose-limiting toxicity (DLT) evaluation was during the first two cycles of chemotherapy. Febrile neutropenia, grade 4 neutropenia, thrombocytopenia or anemia, grade 3 or 4 mucositis, uncontrolled diarrhea (in this study, delayed-onset diarrhea (6), grade 3 or 4 and not controlled by optimal supportive measures, was defined as uncontrolled diarrhea), nonhematologic toxicities (other than controlled nausea or vomiting), and dose delays beyond 2 weeks represented DLTs. No dose modifications were permitted during the first two cycles. For cycles 3 and beyond, irinotecan and 5-FU were reduced by 25% and 50% for grade 3 and 4 toxicities, respectively. Febrile neutropenia required a 50% dose reduction.

Pretreatment and follow-up studies. Before initiation of therapy, all patients had a history and physical examination; assessment of Eastern Cooperative Oncology Group performance status; 12-lead electrocardiogram; determination of tumor measurements with computerized tomographic scans of chest, abdomen, and pelvis; dipstick urinalysis; and routine laboratory studies that included a complete blood count with differential WBC count, electrolytes, urea, creatinine, glucose, total protein, albumin, calcium, phosphate, uric acid, alkaline phosphatase, total and direct bilirubin, and alanine aminotransferase and aspartate aminotransferase levels. History, physical examination, and laboratory tests were repeated on day 1 of each cycle of therapy. Assessment of toxicity and hematology tests were done weekly during each cycle of therapy. Tumor assessments were done after every 8 weeks of therapy, and response was assessed using Response Evaluation Criteria in Solid Tumors (7).

Methodology. For pharmacokinetic measurements, 5 mL blood was drawn into heparinized tubes on day 1, cycle 1, and cycle 2 at the following times, measured from the beginning of the 90-minute infusion: pretreatment (time 0), 0.75, 1.5, 2, 2.5, 3, 4, 6, 8, 10, and 24 hours.

Measurement of irinotecan, SN-38, and SN-38G. These measurements were conducted using a validated high-pressure liquid chromatography method, with fluorescence detection. The method is a modification of that described by Warner and Burke (8). Campothecin was used as internal standard. The ratio of the peak areas for the irinotecan and internal standard and that for SN-38 and internal standard was used for quantitation. This method measures total irinotecan and SN-38. The limits of quantitation for both are 2.5 ng/mL.

Quality assurance was maintained by simultaneously assaying the quality control samples prepared in bulk, before assay validation. Irinotecan and SN-38 from plasma were extracted with acidified methanol. The residue after evaporation of methanol was dissolved in 3% triethylamine acetate (pH 5.5) and acidified methanol (50:50) and injected onto high-pressure liquid chromatography. Separation was carried out on a Waters (Milford, MA) Nova-Pak C18 column equipped with a μBondapak C18 guard column, with mobile phase consisting of 20% acetonitrile and 80% triethylamine acetate solution (pH 5.5). The detection was by fluorescence, with excitation at 370 nm and emission at 510 nm. For measurement of SN-38G, plasma was incubated with 1,000 units of β-glucuronidase for 2 hours at 37°C, before extraction with acidified methanol and high-pressure liquid chromatography. The β-glucuronidase reaction deconjugates the glucuronide moiety from SN-38, thus giving the total SN-38 (SN-38 + SN-38G). The difference between the total and the SN-38 is a measure of SN-38G.

Pharmacokinetic/pharmacodynamic modeling and statistics. A mathematical pharmacokinetic/pharmacodynamic model was developed to characterize drug exposure and its relationship to markers of toxicity. Our group developed pharmacokinetic models for irinotecan and SN-38, which were adapted to this study. Plasma concentrations of irinotecan, SN-38, and SN-38G were analyzed jointly (simultaneously comodeled). Candidate pharmacokinetic models were fit to the data for each individual subject, using a maximum a posteriori Bayesian procedure (Adapt II, release 4), Akaike's Information Criterion for model discrimination, and prior estimates derived from previous irinotecan studies at our institute. Pharmacokinetic variables included area under the plasma concentration-time curve (AUC), C_max, T_max, clearances, and volumes of distribution. Exposure measures assessed included C_max, time above candidate threshold concentrations, AUC_IRINOTECAN, AUC_SN-38, AUC_SN-38G, and various indices of irinotecan/SN-38 relative exposures, including biliary index. Biliary index of SN-38 was defined as the product of the relative area ratio of SN-38 to SN-38G and the total irinotecan AUC (9).

Statistical design. This was a phase I, dose escalation study of irinotecan as part of the combination chemotherapy. Dose escalation was based on 3 + 3 dose escalation rules (10).

There was no intrapatient dose escalation. Primary end point of the study was to determine MTD. Secondary end points included toxicity evaluation, assessment of objective response rates (partial and complete responses), grade 3 or 4 diarrhea, and irinotecan pharmacokinetic studies before and after celecoxib.

Figure 1 and Table 1 show the toxicities in rats treated with irinotecan on the daily × 3 schedule, alone and in combination with celecoxib. The MTD of irinotecan in this model is 100 mg/kg/d. No diarrhea and lethality were observed with irinotecan treated at the MTD. Irinotecan alone at 150 mg/kg produced 75% diarrhea and 50% lethality. With celecoxib, both diarrhea and lethality were reduced to 25%. Irinotecan at 200 mg/kg produced severe diarrhea and lethality in 100% (16 of 16) of treated animals, whereas celecoxib reduced the toxicities of irinotecan at this dose with 25% (4 of 16) diarrhea and 38% (6 of 16) lethality. In the groups receiving the combination of irinotecan with celecoxib, diarrhea was less severe and of shorter duration compared with irinotecan alone. Survival of the groups at 200 mg/kg irinotecan is plotted in Fig. 1B. The median survival of the 16 rats untreated with celecoxib was 7.5 days. However, 10 of the 16 (62.5%) rats treated with irinotecan at 200 mg/kg along with celecoxib had long-term survival (median survival not reached). The difference in survival between the two groups treated with irinotecan at 200 mg/kg due to the addition of celecoxib was marked (P < 0.001, log-rank test). With the weekly × 4 schedule, the MTD of irinotecan is 100 mg/kg/wk. Only toxic doses of irinotecan (150 and 200 mg/kg/wk) were used in combination with celecoxib (30 mg/kg/d × 22, bid) in this schedule. Similar results were achieved (Table 2).

The Ward colorectal carcinoma is relatively resistant to irinotecan treatment. The data in Fig. 1 and Table 1 summarize the antitumor activity of irinotecan alone and in combination with celecoxib with irinotecan given on the daily × 3 schedule and Fig. 2 and Table 2 summarize the results with irinotecan given on the weekly × 4 schedule. Celecoxib (30 mg/kg/d × 7) alone had no antitumor activity. Irinotecan alone had moderate antitumor efficacy with up to 70% of tumor growth inhibition without CR or PR. Response rate as a measure of antitumor activity was zero in all but one of the four treatment groups in Table 1. Only the group of 200 mg/kg irinotecan with celecoxib had responses; the response rate was 75% (50% PR and 25% CR of 16 rats). The presence of synergy was based on the detection of response with the combination of celecoxib with 200 mg/kg irinotecan in the regression model. This was significant at P < 0.001 (exact logistic regression). The data suggest that the daily schedule may be more active than the weekly schedule when irinotecan is combined with celecoxib and that high irinotecan doses are required to achieve tumor responses for Ward colorectal carcinoma.

Thirteen patients consented to participate in the study. One patient withdrew from treatment and was replaced. Thus, 12 patients participated and were evaluable.

Baseline patient characteristics are depicted in Table 3. All but one patient had received prior chemotherapy. Three, six, and three patients were treated at DLs 1, 2, and 3, respectively. There was no DLT at DL 1. At DL 2, one patient had febrile neutropenia and grade 3 diarrhea, which was classified as a DLT. At DL 3, one patient had a grade 4 myocardial infarction, which was a DLT. This patient had an esophageal malignancy, which had previously been treated with chemotherapy. Another patient at DL 3 had grade 3 nausea and grade 3 vomiting, the final DLT. This patient had an unknown primary, which had been treated previously with chemotherapy and radiation. The MTD is DL 2, irinotecan at 200 mg/m², based on the study design.

Adverse events >grade 2 during all treatment cycles are reported in Table 4. For each patient, the maximum grade of an adverse event is depicted. Two cases of grade 4 neutropenia occurred. One patient with rectal cancer at DL 1 had grade 4 neutropenia. As this occurred after two cycles, this event was not considered as a DLT. After four cycles of treatment, this patient was withdrawn from the study due to progressive disease. The second patient had short lived, grade 4 neutropenia, which was not a DLT. No patient suffered from grade 3 or 4 mucositis.

Three grade 4 cardiovascular adverse events occurred in a single patient at DL 3. This patient experienced a grade 4 myocardial infarction that was a DLT; he also developed a grade 4 pulmonary embolism. This event occurred on day 2 of the study regimen, during cycle 1. He made an uneventful recovery and treatment was changed to an alternative regimen. This patient subsequently disclosed a history of 5-FU-induced angina. No grade 4 gastrointestinal adverse event occurred. Only two patients had grade 3 diarrhea. At DL 1, two patients had grade 2 diarrhea. Three of the six patients at DL 2 had grade 2 or 3 diarrhea. None of the three patients at DL 3 reported diarrhea >grade 2. Three patients had DLT and each received only one cycle of treatment.

The mean number of treatment cycles per patient at DL 1, DL 2, and DL 3, respectively, was 5.3, 9.2, and 5.7. At DL 2, a median of 11 treatment cycles was administered. Three of the six patients at DL 2 and one patient at DL 3 reached the maximum of 12 treatment cycles. Reasons for treatment cessation were as follows: progressive disease in three cases, two were withdrawn by either patient or physician choice, and three were removed for toxicity. All but two patients experienced dose modifications or delays. One patient at DL 1 accounted for three dose modifications and three dose delays within four cycles.

No patient had PR or CR. Eight patients had stable disease; five of the six patients at DL 2 experienced SD. The three cases with DLT were not evaluable. The overall median time to treatment failure (time interval between treatment initiation to the cessation) was 17 weeks. The eight patients having best response of SD received a median of 11 cycles (range, 4-12).

Pharmacokinetics of irinotecan, SN-38, and SN-38G on days 1 (before celecoxib) and 14 (after celecoxib) are represented in Fig. 3 and Table 5. As shown, celecoxib had limited effect on the pharmacokinetics of irinotecan, SN-38, and SN-38G in our study. However, due to the limited number of patients enrolled in this study (four patients had pharmacokinetic sampling, three at DL 1 and one at DL 2), potential interaction between celecoxib and the above chemotherapeutic agents cannot be ruled out.

COX-2 overexpression occurs in several cancer types. COX-2 inhibitors interfere with tumorigenesis and apoptosis. Therefore, COX-2 and its gene product may be attractive targets for therapeutic strategies in cancer. Our preclinical, in vivo data from the Ward rat model indicate that higher doses of irinotecan resulted in an incremental antitumor effect and this response was enhanced by the addition of celecoxib. Further, in this model, celecoxib ameliorates irinotecan-induced diarrhea and lethality. The use of a validated preclinical model to evaluate chemotherapy-induced diarrhea along with the statistical power of our observations strengthens the rationale for doing clinical studies with the combination of irinotecan and celecoxib. Kase et al. (11) showed earlier that irinotecan-induced early- and late-onset diarrhea occurs as a consequence of mucosal injury and is accompanied by increased prostaglandin E₂ production. Zweifel et al. (12) noted in a head and neck xenograft model that inhibition of COX-2 by celecoxib resulted in loss of intratumor prostaglandin E₂ levels and reduced tumor growth in a dose-dependent manner. These data formed the basis for our clinical study, which explored the role of celecoxib in mucosal protection and amelioration of diarrhea using a phase I, dose escalation schema. In the present study, diarrhea was a DLT in one case only (at DL 2). Sustained periods of disease stability resulted in patients receiving second-line chemotherapy for advanced esophageal cancer and the majority of these received therapy for >6 months. Oral mucositis >grade 2 did not occur in our study population. No pharmacokinetic interactions occurred between celecoxib and irinotecan or its metabolites, including SN-38 and SN-38G in this patient population.

Others have investigated the antitumor role of COX-2 inhibitors in combination with chemotherapy or radiation. The COX-2 inhibitor nimesulide enhanced the cytotoxicity of several chemotherapeutic agents in non–small cell lung cancer cell lines (13). Celecoxib, when administered as a single agent in escalating doses with radiotherapy, led to significantly lower rates of radiation-induced esophagitis (14). Two phase I clinical studies investigated the combination of celecoxib, docetaxel, and irinotecan for the treatment of patients with advanced solid tumors (15, 16). In both studies, no significant amelioration of diarrhea occurred; however, disease stabilization occurred in several of the treated patients. Pan et al. (17) conducted a phase II clinical trial for patients with advanced colorectal cancer with the IFL regimen in combination with celecoxib and glutamine; toxicities were substantial (45%, grade 3 diarrhea). Amelioration of mucositis was not the primary goal of any of these studies; further, none of these studies accounted for UGT1A1 polymorphism. Patients with certain UGT1A1 polymorphisms have significantly lower SN-38 glucuronidation rates, which predisposes them to diarrhea and neutropenia (4, 18–20). The role of celecoxib in patients who are not heterozygous or homozygous for UGT1A1 mutation is unknown.

There were important differences between our preclinical and clinical studies. In the in vivo studies, celecoxib was administered 24 hours before irinotecan; this was not the case in the clinical study where celecoxib was started on day 2. The rats treated with irinotecan plus celecoxib on a daily × 3 schedule fared better (antitumor activity and diarrhea) than the rats treated on weekly schedule. Amelioration of diarrhea and potentiation of antitumor effect by celecoxib may depend on sequence and frequency of administration of celecoxib and irinotecan, respectively. Further, in our in vivo studies, rats did not receive concurrent 5-FU and leucovorin. The two-weekly FOLFIRI schedule was chosen in our study as this is a common schedule used in the clinic. It remains to be determined if the discordance between the in vivo and clinical study designs resulted in discrepant outcomes. A limitation of our study was that celecoxib dose was not escalated. A recent study investigated the optimal biological dose of celecoxib and reported that 600 mg bid was the preferred dose in combination with erlotinib (21). The validity of this dose in combination with FOLFIRI has not yet been explored.

In conclusion, our study showed that irinotecan doses of 200 mg/m² can be safely administered in the FOLFIRI schedule, when combined with celecoxib. The mucosal protective role of COX-2 inhibitors should be investigated in a prospective, randomized phase II or III clinical trial after excluding cases with UDP-glucuronosyltransferase polymorphism associated with increased irinotecan toxicity.