A Randomized, Placebo-Controlled, Phase II, Presurgical Biomarker Trial of Celecoxib Versus Exemestane in Postmenopausal Breast Cancer Patients

To increase drug development efficiency and assess tumor sensitivity to a given drug, presurgical windows of opportunity trial design have been successfully used. Ki-67 labeling index modulation has been shown to be an appropriate endpoint biomarker for preoperative studies involving hormonal therapies. The posttreatment tumor expression of Ki-67 has been shown to be correlated to disease-free survival (1, 2) and overall survival (3). In addition, Ki-67 analysis on adjacent intraepithelial neoplasia or atypical hyperplasia can be used to explore the potential preventive activity of the tested agent (4, 5).

Exemestane is an irreversible aromatase inhibitor (AI). It binds covalently to the substrate-binding site of aromatase and thereby inactivates the enzyme. This drug is highly specific and inhibits peripheral conversion of androstenedione to estradiol up to 98% (6). The drug is largely used at all stages of breast cancer (7), including the primary prevention setting, where an overall 65% reduction and a 73% reduction of estrogen receptor (ER)-positive breast cancers were shown in the phase III MAP.3 trial (8). Unlike other AIs, the steroidal structure of exemestane suggests that it may have androgenic properties that may counteract the antiestrogenic properties on the bone, sexual function, and other menopausal symptoms (9).

Epidemiologic studies have explored the link between inflammation and cancer, confirming the role of anti-inflammatory drugs in decreasing cancer risk. The inhibition of COX-2, the inducible isoform of cyclooxygenase is one of the pathways involved. Induction of COX-2 by inflammatory stimuli results in the biosynthesis of prostaglandin E2 (PGE2) that control the inflammatory response (10). COX-2 is widely expressed in both invasive and preinvasive breast lesions (11) and PGE2 biosynthesis is considered one check point in the mammary carcinogenesis pathway. Moreover, increased PGE2 levels stimulate aromatase transcription (12). Therefore, NSAIDs and selective anti-COX-2 inhibitors may reduce breast cancer risk through the downregulation of aromatase expression (13). Observational studies, with some inconsistency, have demonstrated an association of NSAID use with reduced breast cancer risk, recurrence, and death (14–16). Among the COX-2 selective inhibitors, celecoxib has a relatively safe profile and has been considered as a possible breast cancer preventive agent (17).

We performed a randomized, placebo-controlled presurgical trial to assess the antiproliferative activity of exemestane and celecoxib in a window-of-opportunity trial in women with ER-positive early breast cancer. The primary outcome was the change in Ki-67 in tumor biopsies before and after 6 weeks of treatment to confirm the antiproliferative effect of exemestane and celecoxib. The expression of ER and progesterone receptor (PgR) was before and after treatment. Secondary endpoints were a panel of circulating biomarkers, including testosterone and sex hormone binding globulin (SHBG), lipid profile, antithrombin III, fibrinogen; C-reactive protein (CRP), C-telopeptide or carboxy-terminal collagen crosslinks (CTX), osteocalcin, and a major urinary metabolite of PGE2 (PGE-M).

We conducted a randomized phase II placebo-controlled trial in postmenopausal women with stage I to II ER-positive breast cancer. The treatment arms were exemestane (25 mg/day), celecoxib (800 mg/day), or placebo for 6 weeks before surgery. Exemestane and placebo were double blinded, whereas celecoxib, due to the different dosage and schedule, was open. The arm ratio was 2:2:1 for exemestane, celecoxib, and placebo, respectively. The duration of the treatment was based on the average waiting list in our institute when the study was conducted. The study (IEO number 162, register number ISRCTN86894592) and all amendments during its conduct were approved by the Institutional Review Board (European Institute of Oncology Ethical Committee). Main eligibility criteria were: postmenopausal histologically confirmed ER-positive primary breast cancer (stage T_1–2 N_0–1, M₀), eligible for surgery; signed informed consent. Women with larger tumors who refused neoadjuvant treatment were also eligible.

Exclusion criteria included: previous treatment for breast cancer including chemotherapy and endocrine therapy, coexisting malignancies diagnosed within 5 years with the exception of basal cell carcinoma or cervical cancer in situ, history of thromboembolic events or other cardiovascular diseases, current anticoagulant therapy, moderate to severe alterations in hematologic profile and hemostasis, dysfunction in renal and hepatic metabolism.

Baseline core biopsies of tumor tissue were collected to confirm tumor characteristics. At baseline visit, blood and urine samples, medical history and anthropometric measurements, concomitant medication, and symptoms were collected. The day before surgery, blood and urine samples, symptoms and toxicity, anthropometric measurements, and concomitant medication were taken. At surgery, a sample of breast cancer tissue was stored. Toxicity was evaluated according to the National Cancer Institute Common Terminology Criteria of Adverse Events (NCI-CTCAE), version 3.0.

Morning fasting blood samples were taken at baseline and on the day before surgery; serum and plasma were separated by 10-minute centrifugation at 1,850 × g and stored at −80°C until assays were performed.

Total cholesterol, HDL-cholesterol (HDL-C), and triglycerides serum levels were determined by a high-sensitivity turbidimetric method with Cobas Integra (Roche Diagnostics), a fully mechanized multichannel analyzer for routine clinical chemistry purposes. Methods were followed according to the specific instructions. LDL-cholesterol was obtained according to the Friedewald formula [LDL-cholesterol = total cholesterol − HDL-cholesterol-(triglycerides/5)].

Serum concentrations of high-sensitive- C-reactive protein (Hs-CRP) were also measured by COBAS INTEGRA 800 according to the manufacturer's instructions. Sensitivity for the hs-CRP assay was 0.1 mg/L and the intra- and interassay coefficients of variation expected was 4% and 6.4%, respectively, for a control sample of 0.423 mg/L.

Plasma fibrinogen and antithrombin III were assayed on plasma citrate samples using the ACL Elite Pro Analyzer (Instrumentation Laboratory). In this assay, clot detection was performed using photo-optical technology.

SHBG, testosterone, CTX, and osteocalcin were evaluated on a Modular E411 immunoanalyzer (Roche Diagnostics). The sensitivity of the assays was 0.0350 nmol/L and 0.025 ng/mL for SHBG and testosterone, 0.01 and 0.50 ng/mL for CTX and osteocalcin, respectively.

With the exception of the lipid profile, fibrinogen and antithrombin III, which were determined on fresh specimens, pretreatment and posttreatment serum samples obtained from each subject were simultaneously assayed on frozen samples to eliminate the effects of interassay variation. In these assays, in addition to the specific control samples provided with an assay kit, an in-house pooled control sample of serum obtained from healthy donors was used to monitor the coefficient of variation between assays.

First morning urine samples (15 mL) from the same patients were obtained under fasting condition at baseline and the day before surgery. Urinary creatinine was determined to normalize urinary biomarker results.

Because of the rapid metabolism of PGE2, the determination of the in vivo biosynthesis was accomplished by the measurement of the concentration of PGE2 metabolites. We measured urinary concentrations of PGE metabolites (PGEM) by a competitive enzyme immunoassay (EIA) kit purchased from Cayman (Cayman Chemical Co.) that converts all major metabolites into a single stable derivative which is easily measurable by EIA. PGE-M levels were normalized to the creatinine level of the sample to account for differences arising from variations in urine concentrations.

Biopsy and surgical specimens were fixed in 10% neutral-buffered formalin for 6 to 8 hours before being embedded in paraffin. Sections (4-μm thick) were cut and stained with hematoxylin and eosin. Consecutive serial sections were used for immunohistochemical determinations. Expressions of ER, PgR, Ki-67, HER2/neu were determined by IHC, as described previously (18). Tumor subtypes were classified by IHC into four categories according to the 2011 St. Gallen criteria (19).

In this phase II presurgical trial, we tested the differences in the changes in Ki-67 by treatment arms. The primary endpoint of the study was the percentage change from baseline to surgery. All randomized subjects were analyzed according to the intention-to-treat.

In a single-factor ANOVA study, a sample size of 125 patients (50, 50, and 25 for the three arms exemestane, celecoxib, and placebo) had 85% power to detect a difference in the means of percentage changes in Ki-67 versus the alternative hypothesis of equal means using an F test with a 0.05 significance level. The size of the variation in the means is represented by a SD of 18. This calculation was made based on the hypothesis of a 20% increase in percentage change of Ki-67 in the placebo arm and a 25% decrease in the treatment arms. The common SD within a group was assumed to be 60. Descriptive statistics of subjects' demographics and tumor characteristics at baseline were presented and differences among arms are assessed with nonparametric Wilcoxon tests, when evaluated as continuous variables, or χ² when presented as categorical variables. We also described median values, and interquartile ranges (IQR), of biomarkers at baseline, at surgery, by treatment arms and compared changes and percentage changes, from baseline to surgery, by treatment arms.

The statistical analysis of the primary and secondary endpoints was based on the comparison of the pre–post treatment changes of Ki-67 LI and the other biomarkers adjusting for baseline level. Two orthogonal contrasts were used to compare biomarker changes among treatment groups: any treatment versus placebo and exemestane versus celecoxib. These contrasts were specified a priori and are consistent with the aims of the study. No adjustment for multiple comparisons on the secondary endpoints was made because they were exploratory and should therefore be considered hypothesis-generating rather than definitive.

The effect of potential covariates, such as age, body mass index (BMI), tumor grade and size, and baseline levels of biomarkers, together with their interactions with treatment, were investigated via the ANCOVA analysis. We verified the normal distribution of residuals of the full model. All statistical tests were two-sided. Analyses were performed using SAS statistical software (version 9.0, SAS Institute Inc).

A total of 125 out of a total of 252 screened postmenopausal patients were recruited from February 2004 to March 2009. They were randomized to either exemestane 25 mg/day, or celecoxib 800 mg/day or placebo (50:50:25 subjects each arm). Twenty-two women were not eligible, 26 refused to participate in the study, and 79 entered in competing trials (Fig. 1). The three arms were well balanced and the main characteristics of the randomized patients are shown in Table 1. Compliance was checked by a self-reported diary and pill count. Overall compliance was very high, 74% of the patients showed a pill intake > 80%.

The pretreatment versus posttreatment changes in Ki-67, ER, and PgR are shown in Table 2. At baseline, the median level of Ki-67 was 22% (IQR 16–27) in the exemestane group and 18% in the celecoxib and placebo arms (IQR 12–22 and 15–25, respectively). After 6 weeks of treatment, a significant absolute reduction in Ki-67 of 10% (IQR −18 to −5), and over a 50 % relative reduction from baseline was observed in the exemestane group, whereas no change was observed in the two other arms (P value = 0.002 for any treatment versus placebo and P value < 0.0001 for exemestane vs. celecoxib). Exemestane also induced a significant reduction in PgR expression from baseline (median change −15% (IQR −52–0) corresponding to a −77% (IQR −0.19 to −1) relative change (P value = 0.002 for treatment vs. placebo and P value < 0.0001 for exemestane vs. celecoxib) and no significant change in ER expression (Table 2).

The levels of circulating biomarkers are shown in Table 3. Patients receiving exemestane showed an increase in testosterone from baseline, whereas SHBG levels significantly decreased (median change from baseline 0.21, IQR 0.12–0.35 and −14.6; IQR −23.1 to −8.6 for testosterone and SHBG, respectively). No change was observed in the other two arms (P < 0.0001 for any treatment vs. placebo and exemestane vs. celecoxib for testosterone; P = 0.0007 and P < 0.0001 for any treatments vs. placebo and. exemestane vs. celecoxib for SHBG, respectively).

Several cardiovascular risk biomarkers were analyzed. They included total and HDL cholesterol, triglycerides, fibrinogen, and antithrombin III. Triglycerides were significantly reduced by exemestane or celecoxib compared with placebo (median change −0.5, IQR −17.5–13.5 and −8; IQR −28–9 for celecoxib and exemestane, respectively; P value = 0.03). Total cholesterol was significantly reduced by exemestane versus celecoxib (P = 0.0006), and HDL-C was reduced by exemestane (P = 0.04 for any treatment vs. placebo and P < 0.0001 for exemestane vs. celecoxib). No changes in fibrinogen, antithrombin III, CRP, and PGE-M were observed.

Osteocalcin did not show any relevant change, wheras CTX was reduced by celecoxib compared with exemestane (P = 0.002).

No grade 3 adverse event was detected after 6 weeks of treatment. In the exemestane arm, 9 subjects (18%) experienced a worsening or de novo mild to moderate hot flashes versus 3 subjects (12%) in the placebo arm and 4 subjects (8%) in the celecoxib group. Six subjects (12%) reported a worsening of night sweating on exemestane versus 2 (8%) in the placebo arm and none in the celecoxib group. Seven subjects (14%) had grade 1 or 2 gastrointestinal toxicity on celecoxib versus 2 (4%) in the exemestane arm and none in the placebo arm. Musculo-skeletal pain was reported in 2 subjects (4%) on exemestane whereas this symptom ameliorated in 4 subjects (8%) in the celecoxib arm.

This trial was designed as a window-of-opportunity study to test the biologic effects of exemestane and celecoxib. The main endpoint was cancer cell proliferation reduction measured by Ki-67 expression. This biomarker together with the secondary endpoints explains some of the mechanisms through which these drugs can act as cancer preventive agents. The relevance of the Ki-67 as surrogate biomarker is documented in several presurgical studies where Ki-67 after presurgical treatment predicts disease-free and overall survival (3, 20, 21). Our results clearly showed a remarkable 10% absolute reduction in Ki-67 expression after exemestane treatment. Ki-67 reduction matched with a remarkable 15% absolute PgR expression reduction, another marker of drug biologic activity (22).

In our previous presurgical trial with different tamoxifen doses (23), the Ki-67 median percent change was much lower compared with that with exemestane (−-54% and 15%, for exemestane and tamoxifen, respectively). Overall, all three registered AIs showed a comparable antiproliferative activity and a greater Ki-67 reduction compared with tamoxifen (24). The IMPACT trial compared anastrozole versus tamoxifen or the combination; the anastrozole arm showed a greater Ki-67 reduction compared with the other two arms (25). The P024 study compared tamoxifen to letrozole with similar results (26). Importantly, all these studies showed a correlation between Ki-67 level after treatment and disease-free survival. Also presurgical exemestane treatment modulates Ki-67 and this correlates with the clinical response (27–29). However, the Ki-67 values are not easily comparable among studies due to the technical variability between different laboratories (30). The clinical relevance of the Ki-67 reduction in the prevention setting is highlighted by the effect of AIs shown in phase III trials, where exemestane and anastrozole reduced breast cancer incidence by 50% to 60% (31, 32) compared with the 40% reduction of SERMs (33).

Consistent with prior studies (34, 35), testosterone levels were increased by exemestane, with a possible greater effect compared with the nonsteroidal AIs. This effect is biologically plausible as AIs inhibit the enzyme which converts androgens to estrogens, even though increased levels of testosterone have not been found in all studies (36). Testosterone is a clear breast cancer risk factor for healthy premenopausal women (37, 38). More recently, it has been shown that testosterone can be a biomarker of recurrence in postmenopausal women (39). Interestingly, in a case–control study nested to the WHEL trial, testosterone and SHBG levels were not associated with increased risk of recurrence, whereas higher serum estrogen concentrations were associated with recurrence risk (40). However, a more recent publication from the WHEL study showed a direct correlation between bioavailable testosterone levels and higher risk of recurrence in postmenopausal women with hot flashes (41). While it is unclear whether testosterone is a breast cancer risk biomarker per se or as a precursor of estrogen (42), it is unknown if the increase in testosterone on exemestane may attenuate its preventive potential long-term.

On the basis of the modulation of circulating biomarkers, exemestane may have a neutral cardiovascular safety/toxicity profile, as in our study it decreased total cholesterol and triglycerides but also, to a lesser extent, HDL-C. AIs have been associated with increased cerebrovascular risk in some studies (43, 44), but not all studies, possibly due to their adverse lipid modulation (45).

The presumed better effect on bone metabolism due to the steroidal structure compared with the other nonsteroidal AIs remains controversial (46, 47). Bone-circulating biomarkers were not modified by short-term exemestane treatment. AIs have clear side effects on bone mineral density and fracture risk which is usually associated with an increased bone resorption and formation markers with a peak at 12 to 18 months (48). However, a reduction in bone mineral density is not always preceded by alterations of bone turnover markers (49).

Our results on Ki-67 are in line with those arising from a similar study in ductal carcinoma in situ (DCIS) treated with exemestane, celecoxib, their combination, or placebo (50). The data were comparable for the exemestane effects on Ki-67 and PgR and no effect by celecoxib. Conversely, in another study in COX2-expressing DCIS, the combination of exemestane and celecoxib showed a possible synergistic effect (51).

Celecoxib alone has been evaluated in previous breast cancer presurgical studies with contradictory results: Martin and colleagues showed mainly null results with 400 mg/day treated for 2 weeks (52), whereas Brandão and colleagues used 400 mg twice a day for 2 to 3 weeks and showed a significant Ki-67 reduction (53). Both studies evaluated a relatively small group of subjects.

Regarding side effects and quality of life, only minor increased climacteric symptoms in the exemestane arm and gastrointestinal discomfort in the celecoxib arm were observed, with no grade 3 adverse events. In the MAP-3 prevention trial, exemestane did not affect quality of life and did not increase fracture risk compared with placebo, notwithstanding a reduction in bone mineral density (8, 49, 54). In the CAAN trial (49), exemestane in combination with celecoxib showed a significantly better quality of life compared with exemestane alone or letrozole. The modest excess of gastrointestinal toxicity in the celecoxib arm should also be interpreted with caution as treatment was unblinded.

The strengths of this study are: the relatively high sample size, the high compliance, and the evaluation of the Ki-67 in a referral laboratory for this methodology (55). It may be considered a weakness that there was a 2:2:1 randomization ratio with a lower power to detect subtle biomarker changes. Also the combination arm was excluded despite the possible interaction between exemestane and celecoxib, because our aim was to develop prevention strategies that were simple and acceptable by high-risk subjects. Results of exploratory analyses on circulating biomarkers should also be considered with caution because the study was designed to investigate differences in Ki-67 between any treatment versus placebo or exemestane versus celecoxib.

The low uptake of breast cancer preventive drugs (56) has led to the use of alternative schedules and doses to minimize toxicity and increase acceptance (57). We are planning a presurgical trial of alternative dosing of exemestane.

In conclusion, our findings indicate that Ki-67 is significantly modulated after short exemestane treatment. Given the known preventive efficacy of exemestane, our findings support the use of Ki-67 as a reliable surrogate biomarker of preventive efficacy. In contrast, there is no clear evidence to support the possible role celecoxib in breast cancer prevention.

No potential conflicts of interest were disclosed.

Conception and design: V. Aristarco, D. Serrano, M. Lazzeroni, G. Pruneri, A. DeCensi, B. Bonanni

Development of methodology: H. Joahnsson, D. Macis, G. Pruneri, B. Bonanni

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): D. Serrano, H. Joahnsson, D. Macis, M. Lazzeroni, I. Feroce, G. Pruneri, A. Toesca

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Gandini, A. Guerrieri-Gonzaga, G. Pruneri

Writing, review, and/or revision of the manuscript: V. Aristarco, D. Serrano, S. Gandini, D. Macis, G. Pruneri, A. DeCensi, B. Bonanni

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): H. Joahnsson, D. Macis, A. Guerrieri-Gonzaga

Study supervision: A. Guerrieri-Gonzaga, A. DeCensi, B. Bonanni

Other (laboratory assays and technical support): V. Aristarco

Other (performed the core biopsy):G. Pagani

Other (breast biopsy):P. Caldarella

B. Bonanni received the financial support by European Institute of Oncology Foundation.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.