Biological Activity of Celecoxib in the Bronchial Epithelium of Current and Former Smokers Free

Non–small cell lung cancer (NSCLC) is the leading cause of death from cancer among both men and women in the United States, accounting for ∼28% of such deaths. Indeed, an estimated 160,000 Americans died of NSCLC in 2007. In recent years, the incidence of NSCLC has begun to decline among men (1). However, smoking-related NSCLC has continued to increase among women, surpassing even breast cancer as the leading cause of cancer death in this group (2). Despite aggressive treatment strategies, the 5-year survival rate for NSCLC remains only ∼15% (1). These grim facts underscore the urgent need for a change in our approach to NSCLC.

Smoking prevention and cessation have been emphasized as ways to reduce deaths from cancer. Despite the reduction in NSCLC risk observed with smoking cessation, however, several studies have shown that former smokers still have a higher NSCLC risk than nonsmokers have (3–7) and consequently account for a large proportion of NSCLC patients in this country. Chemoprevention strategies, especially for high-risk populations such as former smokers, are appropriate in NSCLC. However, large-scale NSCLC chemoprevention trials, including the Alpha-Tocopherol Beta-Carotene trial, Beta-Carotene and Retinol Efficacy Trial, and Lung Intergroup Trial, have yet to show that any agent can reduce lung cancer risk (8–11).

One of the changes identified in premalignant bronchial tissues that has potential therapeutic significance is an increase in expression of cyclooxygenase-2 (COX-2). COX-2 converts arachidonic acid to prostaglandin H₂, a precursor of prostaglandin E₂ that has been implicated in a variety of biochemical processes, required for cell proliferation and survival, and whose expression increases in response to growth factors, oncogenes, and carcinogens (12–18). COX-2 has been extensively studied in epithelial tumors, including colorectal cancer and NSCLC (19–22). COX-2 overexpression has prognostic value, predicting a worse outcome in NSCLC patients with stage I disease whose tumors have been surgically resected (23, 24) and thus suggesting that COX-2 is an important biological determinant in NSCLC. Tellingly, COX-2 expression is higher in bronchial premalignant lesions than in adjacent normal lung tissue (22, 25), raising the possibility that COX-2 promotes malignant progression in the lung. Moreover, COX-2 inhibitors have shown efficacy as NSCLC chemopreventive agents in preclinical studies, reducing the size and number of carcinogen-induced NSCLC tumors in mice (26). These findings provide a compelling rationale to investigate the activity of COX-2 inhibitors as chemopreventive agents for lung cancer.

In this study, our goal was to determine whether a 6-month treatment with celecoxib, a COX-2 inhibitor, would be safe and have biological activity in the lungs of current and former smokers. We did a randomized, placebo-controlled study to examine the toxicity and efficacy of celecoxib; bronchial epithelial cell proliferation, as measured by the Ki-67 labeling index after 3 months, was the primary end point. We chose this primary end point on the basis of evidence that bronchial premalignant lesions increase epithelial cellular proliferation and that COX-2 promotes cellular proliferation and survival (27–29).

For the present study, we recruited current smokers (those actively smoking or those who had quit within the previous 12 mo) and former smokers (those who had quit at least 12 mo before study entry) who had at least a 20 pack-year history of smoking (pack-year = number of packs of cigarettes per day times number of years smoked). Patients could have had prior stage I NSCLC or stage I or II laryngeal cancer but had to have been free of disease for at least 6 mo before study entry. Other exclusion criteria included the chronic use of steroids, the use of H₂ blockers for active ulcers, the use of nonsteroidal anti-inflammatory drugs other than low-dose aspirin of ≤81 mg/d, and a history of stroke, uncontrolled hypertension, and/or angina pectoris. Patients were recruited through local community groups, health fairs, and advertisements distributed to referring practitioners and patients at The University of Texas M.D. Anderson Cancer Center. The study was approved by the Institutional Review Board and by the U.S. Department of Health and Human Services. All patients provided written informed consent before entering the study.

The clinical study design was a four-arm, double-blind, placebo-controlled, randomized study to evaluate the biological effects of celecoxib as a chemopreventive agent in current and former smokers. The primary end point was modulation of Ki-67 in the bronchial epithelium after a 3-mo period of treatment. Patients were treated for up to 6 mo and were randomized onto one of four treatment arms: celecoxib daily for 3 mo, then placebo daily for 3 mo (CCX + PCB); celecoxib daily for 3 mo, then celecoxib daily for 3 mo (CCX + CCX); placebo daily for 3 mo, then celecoxib daily for 3 mo (PCB + CCX); and placebo daily for 3 mo, then placebo daily for 3 mo (PCB + PCB). The research pharmacy randomly assigned each patient to one of the four treatment arms and recorded this assignment by using a computer-generated treatment code that was available only to the pharmacist. Pfizer Corp. provided the 200-mg celecoxib capsules and the matching placebo capsules.

After a bronchoscopy at 3 mo, patients received treatment for an additional 3 mo. A bronchoscopy was then done at the 6-mo time point.

On completing informed consent and enrollment, patients were screened with a chest X-ray and bronchoscopy, which included bronchial washings, brushings, and biopsies from six predetermined sites (carina, right lower, middle, and upper lobes and left lower and upper lobe regions) as well as from any abnormal sites suspicious for cancer. Metaplasia indices (MI) were calculated from the biopsies done at the predetermined sites. The presence of dysplasia was confirmed by histologic evaluation of all biopsy samples. Patients with severe dysplasia at initial or subsequent bronchoscopy were strongly encouraged to undergo additional bronchoscopies at 6 mo.

Patients were stratified for randomization according to smoking history (current versus former), prior cancer (prior versus no prior), and MI (<15% versus ≥15% and/or dysplasia). Toxicity was monitored using the National Cancer Institute Common Toxicity Criteria 2.0, and patients who experienced grade 2 or higher toxicity had their dose reduced. Clinic visits occurred before treatment and during treatment at 1-mo intervals. A complete physical examination, which included asking about the patient's relevant medical history and history of tobacco and alcohol exposure, was done at each clinic visit. Patients were referred to smoking cessation programs on request.

In the original study design, celecoxib was to be administered at 200 mg twice daily. At a scheduled External Advisory Board meeting, the committee raised the concern that in a recently completed colon polyp prevention study (20), a 100-mg dose of celecoxib did not differ from placebo in terms of polyp reduction. On the basis of this updated data, the External Advisory Board recommended a higher dose of celecoxib (400 mg). Therefore, the starting dose level, effective December 2003 (after 81 patients had been enrolled at the low-dose celecoxib level), was set at 400 mg for new patients randomized to receive celecoxib. The first subject was randomized using the new schedule on January 23, 2004.

On December 17, 2004, reports of cardiovascular toxicity in other COX-2 inhibitor trials were released (30–35). At that point, a total of 150 patients had been registered on the current study and 143 had been randomized to treatment. New subject entry and celecoxib treatment were put on hold by the principal investigator of the study and The M.D. Anderson Cancer Center Institutional Review Board. During the subsequent months, efforts were made to follow-up with participants, audit clinical data and the laboratory specimen inventories, modify the eligibility criteria to exclude patients with preexisting cardiovascular conditions, and include additional procedures to screen and monitor for cardiovascular toxicities. The protocol was amended to address cardiovascular safety issues to ensure the safety of study patients. The M.D. Anderson Institutional Review Board approved the amended protocol, the study was reactivated in April 2005, and we began registering new patients; seven of the patients whose treatment was put on hold reentered the trial. We stopped patient accrual for this trial as of January 2007.

Per the protocol, patients underwent bronchoscopies at the time of enrollment before randomization. All evaluable study patients then had repeat bronchoscopies with biopsies, brushings, and washes at the completion of the first stage of treatment (3 mo) and again at 6 mo. Biopsy, brushing, and wash samples were obtained from the same predetermined sites sampled in the initial bronchoscopy. As noted, these biopsy specimens were taken at six predetermined sites in the bronchial tree: the main carina, the bifurcation of the right upper lobe and the main stem bronchus, the bifurcation of the right middle lobe and right lower lobe, the bifurcation of the left upper lobe and lingula, the medial bronchus of the right lower lobe, and the anterior bronchus of the left lower lobe. We fixed the biopsy specimens in 10% buffered formalin, embedded them in paraffin, and sectioned them. The first two 4-μm tissue sections from each biopsy site were stained with H&E and evaluated for the presence of squamous metaplasia and dysplasia. We did histologic assessments to determine whether the MI had changed during the 3-mo period. The MI was calculated as the percentage of biopsy sections exhibiting squamous metaplasia out of the total number of sections examined. A single pathologist (X.T.) who was blinded to the study treatment served as the reference pathologist.

We cytologically analyzed sputum samples acquired by sputum induction from all patients before treatment and after 3 and 6 mo of treatment. Additionally, we did buccal brushings on all patients before treatment and after 3 and 6 mo of treatment to look for evidence of tobacco-induced histologic and genetic alterations.

We calculated the fraction of Ki-67–positive cells in the bronchial epithelium, including the basal, parabasal, and superficial layers of the biopsy specimens. Ki-67 labeling indices were expressed as the percentage of cells with positive nuclear staining, as detailed in our prior reports (27, 28). Slides that lacked evaluable epithelium were excluded from the analyses. Ki-67 labeling indices were analyzed on a per-biopsy-site basis and on a per-subject basis (the average of all biopsy sites that could be evaluated from a participant at a particular time point).

The immunohistochemical analysis was done as follows: one 4-μm tissue section was deparaffinized in xylene and rehydrated through a series of alcohols. Peroxide blocking was done by immersing the section in 3% hydrogen peroxide in methanol for 15 min. Antigen retrieval was accomplished by placing slides in a steamer for 10 min with 10 mmol/L sodium citrate (pH 6.0) and washing them in Tris buffer. The slides were then blocked in 10% fetal bovine serum for 35 min. The Ki-67 antibody used was MIB-1 (DAKO), and incubation occurred at room temperature at 1:200 dilution for 65 min. Secondary antibody was provided and detection was done using the Envision Link+ kit (DAKO) for 30 min. Diaminobenzidine chromogen was applied for 5 min. The slides were then counterstained with hematoxylin and topped with a coverslip. We used NSCLC cell line pallet sections that had been formalin fixed and paraffin embedded and that evidenced confirmed antigen expressions as the control cell lines.

This study was designed as a randomized, double-blinded, placebo-controlled trial to evaluate the efficacy and toxicity of celecoxib as a chemopreventive agent in current and former smokers. The planned duration of treatment was a total of 6 mo. The primary end point of the study was modulation of Ki-67, measured after a 3-mo treatment intervention. The secondary end point of the study was the change in Ki-67 labeling at 6 mo. The stratified Z test was applied for comparing the reduction of Ki-67 from baseline to 3 mo between the treatment and placebo groups. The target number of randomized and evaluable patients was 182, which would require a total of 216 randomized patients, allowing for a 15% dropout rate. The study design had at least 80% power to detect a 1.2% difference in the reduction of Ki-67 between the COX-2 inhibitor and placebo, with a two-sided 5% level of significance.

Summary statistics, including frequency, tabulation, mean (and SD), and median (and range), were used to characterize the distribution of Ki-67 labeling indices in the basal layer, parabasal layer, and all layers. The mean Ki-67 index across all six potential biopsy sites was computed with the patient used as the analysis unit. The Wilcoxon rank sum test was used to compare continuous variables between two groups. The Kruskal-Wallis test was used to compare continuous variables among three groups. The χ² test or the Fisher's exact test was used to test the statistical significance of the association between two categorical variables. The Wilcoxon signed-rank test was used to test changes in Ki-67 labeling indices by patient before and after treatment within each treatment group. To increase the efficiency of the statistical analysis, we also used the biopsy site as a unit of analysis under the assumption that the site was nested within the patient.

For these analyses, we used a mixed-effect model to test the effect of treatment with celecoxib against placebo on Ki-67 labeling indices, adjusting for covariates that affect Ki-67 levels, such as number of years since smoking cessation (in categories), squamous metaplasia (presence or absence), treatment arm, and time point (0 or 3 mo). When the mixed-effect model was used, a logarithmic transformation was applied to Ki-67 labeling indices to satisfy the Gaussian distribution assumption. All statistical tests were two-sided, with a 5% type I error rate. Statistical analysis was done using standard statistical software, including SAS release 9.1 (SAS Institute) and S-Plus version 7 (Insightful, Inc.).

From November 2001 to September 2006, a total of 212 patients registered onto the study, with 204 patients randomized to treatment with either agent or placebo. Eight patients were not randomized to treatment arms: two patients because they declined bronchoscopies, two patients because they had concurrent medical conditions, and four patients because of the temporary protocol suspension on December 17, 2004.

Of the 204 patients randomized to study sections, 127 patients (61 receiving low-dose celecoxib and 66 receiving high-dose celecoxib) received baseline and 3-month bronchoscopies and thus had data evaluable for the primary end point analysis (Fig. 1). There were 104 patients who received all three (the baseline, 3-month, and 6-month) bronchoscopies. The characteristics of the patients who were randomized to study sections are listed in Table 1. Although the treatment groups generally had similar characteristics, there were fewer women in the arm treated with PCB + CCX (P = 0.52), no Hispanic patients in the CCX + PCB arm (P = 0.06), and less dysplasia at baseline in the CCX + PCB and PCB + CCX arms (P = 0.10).

More patients than expected dropped out of the study because of the temporary protocol suspension, which may explain why the accrual goal of 182 patients with evaluable data was not reached. Common reasons for study dropout in the low-dose celecoxib (200 mg) group included personal reasons (10 patients), being lost to follow-up (8 patients), and concurrent medical conditions (5 patients).

During the protocol suspension, treatment was suspended for 29 patients, all of whom subsequently left the study. Four patients were randomized but never started the study drug because of safety concerns.

Common reasons for dropout in the high-dose celecoxib group (400 mg) included nonadherence as judged by pill counts (12 participants), having or developing concurrent medical conditions (10 participants), and personal reasons (5 participants).

Patients were stratified into statistical groups according to smoking status (current or former smokers). To monitor smoking status, patients were asked at each visit whether they were actively smoking, and serum cotinine levels were measured. In general, serum cotinine values agreed with the patient reports, but Fig. 2 shows two patients who reported having stopped smoking but had cotinine levels >20 ng/mL at baseline. An additional two former smokers admitted to resuming smoking during the study.

Pill counts were done on a monthly basis to measure treatment adherence, which was excellent. Based on pill counts over the first 3 months of treatment, participants enrolled in the PCB, CCX0, and CCX1 arms took 93.7% (±15.2), 92.1% (±11.78), and 94.5% (±15.3) of the prescribed doses, respectively. Comparable results were observed at 6 months of follow-up (data not shown). There were no differences in adherence levels between the treatment groups.

Of the 204 patients who were randomized to study sections, 92 experienced at least one toxic effect, and a total of 196 toxicity episodes were reported. Fifty-eight patients experienced grade 1 toxicities, 28 patients experienced grade 2 toxicities, and 6 patients reported grade 3 to 4 toxicities (confusion, thrombosis, hyperglycemia, allergic reaction, hypertension, and nausea/abdominal pain), but only hyperglycemia, hypertension, and nausea/abdominal pain were considered to be possibly treatment related (Table 2). No cardiac toxicities were observed in the study. According to protocol guidelines, patients who experienced toxicities of grade 2 or greater had their dose levels reduced to the −1 dose level (100 mg).

Serum celecoxib levels were collected via blood draw and measured at baseline and at defined treatment time points (1, 3, 4, and 6 months) before the patient taking the morning dose. At 3 months, mean celecoxib levels were generally in the low micromolar range (Table 3). Serum celecoxib levels were dose dependent in current smokers (2.71 ± 1.34 with high dose versus 2.11 ± 3.40 with low dose; P = 0.004, Wilcoxon rank sum test) but not in former smokers (Table 3). On the basis of findings from an in vitro study (29), low-micromolar celecoxib levels would be sufficient to have biological effects on NSCLC cells.

A total of 212 patients underwent at least one bronchoscopic procedure each, adding up to a sum of 443 bronchoscopic procedures generating 2,658 biopsy samples. Among them, 1,272 biopsy samples were done at baseline, 762 at 3 months of time, and 624 at 6 months of time. Eighteen baseline biopsy samples were inadequate for histologic interpretation. Of the remaining 1,254 baseline samples, 1,086 (86.6%) had normal histology, 152 (12.1%) had squamous metaplasia, and 16 (1.3%) had dysplasia. Squamous metaplasia or dysplasia was detected in 15.5% (148 of 958) of the samples obtained from current smokers and in 5.7% (14 of 248) of the samples obtained from former smokers. The corresponding MI was higher in current smokers [15.2 (±20.3), n = 162] than in former smokers [5.6 (±10.2), n = 42; P = 0.004, Wilcoxon rank sum test; Fig. 3]. There were no differences in MI values among the treatment groups (Table 4).

The primary end point of the study was modulation of the Ki-67 index from the baseline level after 3 months of treatment. Ki-67 values were measurable in 2,202 biopsy samples (1,069 at baseline, 627 at 3 months of time, and 506 at 6 months of time) obtained from 200 patients randomized to one of the four study groups (Table 5). Wilcoxon rank sum test shows that baseline Ki-67 expression was significantly higher in current than in former smokers among all epithelial layers (6.15 ± 6.01% versus 3.86 ± 5.56%; P = 0.002), the basal layer (6.07 ± 6.77% versus 3.49 ± 4.69%; P = 0.003), and the parabasal layer (9.23 ± 10.42% versus 6.31 ± 12.43%; P = 0.009). Other variables that affected Ki-67 labeling were the presence of squamous metaplasia (P < 0.0001) and the number of quit-years (1 to <5, P = 0.0004; >5, P < 0.0001). We first examined the effect of celecoxib treatment on Ki-67 labeling in all epithelial layers, which was the primary study end point, by combining the low- and high-dose treatment cohorts. Mixed-model analysis revealed that Ki-67 labeling was not significantly different between the celecoxib and placebo groups (P = 0.12). However, although the effect of low-dose treatment was not significant (P = 0.79), 3 months of high-dose treatment decreased Ki-67 labeling in all epithelial layers in both former smokers (3.85% decrease) and current smokers (1.10% decrease), which was a significantly greater reduction in both groups (P = 0.02, mixed-model analysis) than that in the placebo group after adjusting for metaplasia and smoking status (Table 6; Fig. 4A). This treatment effect persisted at the 6-month time point (Fig. 4B), which further supports the idea that there was a biological effect resulting from high-dose treatment. Additional analysis was then done to examine the effect of high-dose treatment on specific epithelial layers. Although changes in the parabasal layer did not reach significance, Ki-67 labeling decreased in the basal layer by 4.14% in former smokers and 1.41% in current smokers, which was a significantly greater reduction than that observed in the placebo arm (P = 0.008, mixed-model analysis).

In this first-ever randomized clinical trial of a 6-month celecoxib regimen in current and former smokers, we found that celecoxib is safe to administer and biologically active in the bronchial epithelium. The effects of treatment on the primary end point, bronchial epithelial proliferation after 3 months of time, are noteworthy given that the participant accrual goal was not reached. Moreover, the biological activity and safety of celecoxib in this cohort warrant additional studies on the efficacy of celecoxib in NSCLC chemoprevention.

Problems encountered during the conduct of this trial highlight several important feasibility issues in planning lung chemoprevention studies. The unanticipated cardiac toxicities reported in large trials examining the efficacy of celecoxib and other nonsteroidal anti-inflammatory drugs in colon cancer chemoprevention (30–35) negatively affected the conduct of this study in several respects. First, participant accrual was interrupted for 6 months. Second, data from patients who had been actively receiving treatment at the time of protocol suspension were deemed inevaluable due to early treatment cessation. Third, patient accrual after the trial reopened proceeded at a slower rate than it had before trial suspension, which suggests that the negative publicity associated with the cardiac toxicity reports adversely affected patient accrual. In fact, we did not observe any cardiovascular toxicity in this cohort. This may have been related to the short duration of celecoxib treatment in this study relative to that of the trials reporting these toxicities, which required treatments of more than 12 months of duration (20, 30, 36–38).

The findings reported here on Ki-67 labeling in the bronchial epithelium are noteworthy for several reasons. First, Ki-67 labeling decreased in participants treated with high-dose but not low-dose celecoxib. Similarly, celecoxib is more efficacious in colon cancer chemoprevention when administered at 400 mg twice daily than at 200 mg twice daily (20, 30, 38). These findings suggest that, although it may have less treatment-related toxicity, low-dose (200 mg) celecoxib has no efficacy in NSCLC chemoprevention and argue against such a trial design.

Second, Ki-67 levels decreased more prominently in former smokers than in current smokers, especially in those patients who completed baseline and 3-month bronchoscopies (Fig. 4A). This effect was also observed in patients who completed baseline, 3-month, and 6-month bronchoscopies (Fig. 4B) and received high-dose celecoxib. It is important to note that these are subset analyses and the number of patients is low, especially in the former smokers. Serum celecoxib levels did not differ in current versus former smokers treated with high-dose celecoxib (data not shown), but detailed pharmacokinetic studies were not done, so we cannot exclude the possibility that the pharmacokinetics of celecoxib contributed to this outcome. Current and former smokers may differ with respect to the role that COX-2 plays in maintaining bronchial epithelial proliferation. In fact, other studies have reported differences between these two groups with respect to bronchial epithelial biology (28, 39, 40).

Third, the reduction in Ki-67 labeling was not accompanied by a decrease in MI, and the effect of celecoxib on Ki-67 did not vary on the basis of histology, indicating that the decrease in Ki-67 was not due to a reduction in bronchial metaplasia, which has been reported to increase Ki-67 labeling (28, 41–43). Dysplasia was uncommon in this cohort, so no conclusions can be made about the effect of celecoxib on this histologic abnormality.

Fourth, Ki-67 decreased more prominently in the basal layer than it did in the parabasal layer, a strikingly different finding from those reported in chemoprevention studies using retinoids, which reduce bronchial metaplasia and are active primarily in the parabasal layer of the bronchial epithelium (27, 28, 44). Collectively, these findings suggest that the basal and parabasal compartments of the bronchial epithelium are biologically distinct, which is consistent with evidence that cells in the basal layer have a low proliferation rate, express progenitor cell markers, and have multipotent differentiation potential (45), whereas cells in the parabasal layer have a higher proliferation rate and have undergone differentiation into mucous-secreting and other epithelial cell types.

Progress in the field of NSCLC chemoprevention research will require the ability to identify individuals at high risk for the development of NSCLC, a way to isolate premalignant bronchial epithelial cells in danger of malignant progression, and a method to elucidate the mechanisms by which these premalignant cells maintain their proliferation and survival. With respect to the latter, findings presented here and elsewhere (46) raise the possibility that COX-2 is one mediator of bronchial epithelial proliferation in current and former smokers. Additional studies are warranted to examine the importance of COX-2 in NSCLC development and explore NSCLC prevention with COX-2 inhibitors.

No potential conflicts of interest were disclosed.

Grant Support: National Cancer Institute grants CA 091844 and CA 16672 (J.M. Kurie), Department of Defense grant W81XWH-04-0142 (W.K. Hong), and Pfizer Pharmaceuticals.