The objective of this study was to determine whether cyclooxygenase-2 (COX-2) overexpression was an indicator of prognosis in patients with International Federation of Gynecology and Obstetrics (FIGO) Stage IIB uterine cervical carcinoma who underwent radiation and concurrent chemotherapy.
Seventy-five patients with FIGO Stage IIB squamous cell carcinoma (SCC) of the uterine cervix who were treated with radiotherapy and concurrent chemotherapy between 1991 and 1996 were divided into two groups according to their COX-2 level in an immunohistochemical study: the COX-2 negative group (n = 54 patients) and the COX-2 positive group (n = 21 patients). The clinicopathologic features, patterns of treatment failure, and survival data for patients in the COX-2 positive group were compared with data from the patients in the COX-2 negative group. Univariate and multivariate analyses were performed to determine the prognostic factors that influenced patient survival.
In the immunohistochemical study, COX-2 overexpression was observed in approximately 30% of patients with FIGO Stage IIB SCC of the uterine cervix. With delayed regression to the initial treatment, the treatment failure rate of patients in the COX-2 positive group was much higher compared with the treatment failure rate of patients in the COX-2 negative group. The higher incidence of central failure and lymph node failure for patients in the COX-2 positive group was statistically significant (48% for the COX-2 positive group vs. 13% for the COX-2 negative group). However, there was no difference in the incidence of hematogenous metastases between the two groups (5% for the COX-2 positive group vs. 7% for the COX-2 negative group). In addition, increased COX-2 expression in tumor cells also was correlated with a shorter interval to tumor recurrence (median interval to recurrence, 9 months in the COX-2 positive group vs. 26 months in the COX-2 negative group). Compared with patients in the COX-2 negative group, patients in the COX-2 positive group had lower overall actuarial and disease free survival rates (overall 5-year actuarial survival rates: 56% for the COX-2 positive group vs. 94% for the COX-2 negative group; P = 0.003). Univariate and multivariate analyses showed that COX-2 overexpression was an independent prognostic factor that surpassed other well-known clinicopathologic parameters.
Despite the fact that carcinoma of the uterine cervix has a tendency to be localized to the pelvis along with an orderly lymphatic spread to the pelvic, para-aortic, and supraclavicular lymph nodes,1 the tumor behavior can be difficult to predict. Although most patients with uterine cervical carcinoma respond relatively well to standard treatment, some tumors are biologically more aggressive or are more refractory to treatment than others. More often, this aggressiveness does not correlate entirely with well-known clinical parameters, such as International Federation of Gynecology and Obstetrics (FIGO) staging, tumor volume, lymph node metastasis, or treatment modalities. Because the success of any treatment depends on a valid prediction of the prognosis, additional molecular targets are needed to assess tumor behavior more adequately for each patient and to individually tailor new strategies for patients with molecular risk factors.2 Among a number of new biologic markers that have been investigated during the last decade, a particular focus has been to identify the significance of cyclooxygenase (COX) as a prognostic indicator for various types of malignancies,3–10 including carcinoma of the uterine cervix.11, 12
It is well known that COX-2 is a key enzyme that catalyzes the conversion of prostaglandins and other eicosanoids from arachidonic acid. Two isoforms of COX have been characterized: COX-1 is expressed constitutively in most eukaryotic cells and is believed to generate prostaglandins for normal physiologic functions, whereas COX-2 can undergo rapid induction in response to a variety of stimuli, including tumor promoters, cytokines, and growth factors.13–15 Several studies have shown that COX-2 is up-regulated in many experimental murine tumor models and human malignancies.14, 16–19 It promotes carcinogenesis, tumor proliferation and growth, and the spread of disease by mediating the pathologic process affecting mitogenesis, cellular adhesion, and immune surveillance.13–17 It also was found that COX-2 overexpression inhibited apoptosis and stimulated tumor angiogenesis.14, 15, 20–24 Simultaneously, elevated COX-2 levels are a known prognostic factor for patients with several types of malignancies, such as human glioma,3 lung carcinoma,4 breast carcinoma,5 esophageal carcinoma,6 gastric carcinoma,7 and particularly colorectal carcinoma.8–10 Nonetheless, the prognostic significance of COX-2 overexpression has not been studied extensively for uterine cervical malignancies. To the authors' knowledge, there are only two clinical studies of uterine cervical malignancies that indicated an aggressive tumor behavior or unfavorable prognosis for patients with tumors that overexpressed COX-2.11, 12
Although both reports identified patients with a poor prognosis as patient cohorts with COX-2 overexpression who had undergone either surgery11 or radiotherapy,12 the studies only included a small number of patients. Moreover, one cohort was comprised a heterogeneous group of patients with a wide range of FIGO disease stages.12 Even though radiotherapy has long been accepted as a standard treatment for patients with FIGO Stage IIB disease, a few randomized clinical trials of radiotherapy and concurrent chemotherapy recently have conferred meaningful benefits for patients with a certain unfavorable prognostic factor.25, 26 However, it remains to be investigated whether COX-2 expression has a prognostic value in patients who received radiotherapy and concurrent chemotherapy. The objective of this study was to determine whether COX-2 overexpression had any influence on the prognosis of patients with FIGO Stage IIB carcinoma of the uterine cervix who were treated with radiotherapy and concurrent chemotherapy.
MATERIALS AND METHODS
Patients and Treatment Protocol
Seventy-seven patients with FIGO Stage IIB invasive squamous cell carcinoma (SCC) of the uterine cervix who were treated with radiotherapy and concurrent chemotherapy at the Department of Radiation Oncology, Yonsei Cancer Center, Yonsei University College of Medicine (Seoul, Korea) between 1991 and 1996 were analyzed. Among them, 2 patients were excluded from the analysis, because their paraffin embedded tissue blocks were unavailable, leaving 75 patients who were eligible for the study. The clinical staging and histologic classification of uterine cervical carcinoma for each patient were based on the FIGO and World Health Organization classification systems. Without exception, all patients had received pretreatment computed tomography (CT) scans or magnetic resonance imaging (MRI). To make the patient population more homogeneous, all patients with FIGO Stage IIB small cell carcinoma or adenocarcinoma were excluded. Patients who were treated with radiotherapy alone or with neoadjuvant chemotherapy and radiation were also excluded. The eligible criteria for radiotherapy and concurrent chemotherapy included patients with high-risk factors, such as a larger tumor volume or lymph node metastases, on an abdominopelvic CT scan or MRI. Details of the treatment protocol have been described elsewhere.27 Briefly, the patients received three cycles of chemotherapy given concomitantly during Week 1, Week 4, and Week 7 of radiotherapy. The chemotherapy regimens consisted of cisplatin and 5-fluorouracil (5-FU). Cisplatin was given as a daily bolus dose of 100 mg/m2, and 5-FU was administered as a 24-hour continuous infusion at a dose of 1000 mg/m2 per day for 5 days. Radiotherapy was delivered with a combination of external irradiation and high-dose-rate intracavitary radiation by a remote afterloading system using iridium 192 sources (Gamma-Med II). External whole pelvis irradiation was performed with a dose of 1.8 grays (Gy) per fraction 5 times per week to a midline dose of 27.0–36.0 Gy. This was followed by high-dose-rate intracavitary radiation with 6 insertions (twice per week) with a fractional dose of 5.0 Gy to a total dose of 30.0 Gy at point A. After high-dose-rate intracavitary radiation, the patients received a second course of external irradiation with central shielding up to a total external dose of 45.0–50.4 Gy. After completing the treatment, all patients were followed every 3 months for at least 5 years.
Tissue Array Block
The recipient blocks were made with purified agar in 3.8 cm × 2.2 cm × 0.5 cm frames. Holes measuring 2 mm size were made on the recipient blocks by a core needle, and the agar core was discarded. The donor blocks were prepared after a thorough evaluation of the hematoxylin and eosin-stained slides. Representative tumor tissues caught from the matching donor blocks were transplanted to the recipient blocks using a 2-mm core needle. The recipient blocks were framed in the mold, which was used to frame the conventional paraffin blocks. Subsequently, paraffin was added to the frame. Consecutive 4-μm-thick sections were cut from the recipient blocks using an adhesive-coated slide system (Instrumedics Inc., NJ) to support the cohesion of the 2-mm array elements on the glass.
Four-micrometer-thick tissue sections were cut from the formalin fixed, paraffin embedded tissue blocks; dewaxed in xylene; rehydrated through graded ethanol solutions; rinsed in phosphate buffered saline (PBS) for 5 minutes; and then immersed in 0.3% hydrogen peroxide in methanol for 30 minutes to block the endogenous peroxidase. For antigen retrieval, the sections were microwaved in a 0.01 mol/L sodium citrate-buffered saline, pH 6.0, for 30 minutes at 95 °C. The slides were then rinsed in PBS for 5 minutes and blocked with a solution of 10% normal rabbit serum in PBS at room temperature for 10 minutes. They were then incubated at 4 °C overnight with the COX-1 and COX-2 primary antibodies from a Cayman (Ann Arbor, MI) at dilutions of 1:800 and 1:1500, respectively. Each antibody was preheated by soaking in a citrate buffer, pH 6.0, based on the unmasking effect of the masked epitope using microwaves for 10 minutes. Nonspecific mouse immunoglobulin G1 monoclonal antibodies (1 mg/mL; 03001 D; PharMingen) were used as a negative control. Tissues were incubated with biotinylated horse antimouse secondary antibodies at a 1:500 dilution (Vector Laboratories, Burlingame, CA), followed by extensive washes in avidin-biotin peroxidase complex at 1:25 dilution. Diaminobenzene was used as the chromogen, and hematoxylin was used as the nuclear counterstain. The cytoplasmic immunoreactivities were classified as a continuum from the undetected level (0%) to diffuse and homogeneous, strong staining (100%). Semiquantitative, three-tier grading was based on the percentage of tumor cells that stained positive in the whole tumor boundary: 0–1, minimal (< 10%); 2, moderately stained (< 50%); and 3, markedly stained (≥ 50%). If the distributions of immunoreactivity were ≥ 50%, then they were classified as the overexpression groups. To rule out the possibility of interpersonal bias, the results were interpreted by only one investigator who was blinded to the clinical outcome.
The clinical profiles and patterns of treatment failure in both groups were compared using a chi-square test. The overall actuarial survival and disease free survival rates were calculated using the Kaplan–Meier method, and the log-rank test was used to compare the rates between the various COX-2 expression groups. Univariate analysis was used to define the prognostic factors that influenced survival. The relative importance of the covariates in determining prognostic factors also was assessed by using a multivariate Cox proportional hazards model. P values ≤ 0.05 were considered statistically significant.
The patterns of staining for both the COX-1 enzyme and the COX-2 enzyme exhibited a marked intratumoral heterogeneity in both intensity and distribution, ranging from tumors with few weakly positive cells to tumors with apparent overexpression. COX-1 immunoreactivity was localized to the adjacent normal cells, whereas focal or diffuse COX-2 immunoreactivity was detected only in tumor cells. COX-1 staining was observed in both normal epithelial cells and stromal cells and also occasionally was observed in fibroblasts, vascular endothelial cells, and vascular smooth cells. By contrast, COX-2 staining was observed in epithelial cells and was observed rarely in stromal cells. Within tumor cells, patterns of COX-2 staining were mostly diffuse cytoplasmic immunoreactivity, although perinuclear immunoreactivity also was observed in some tumor cells (Fig. 1). Although a strong correlation was observed between the staining intensity and distribution (data not shown), tumor sections were grouped according to the staining distribution. With the extent of the distribution of staining > 50% classified as positive immunoreactivity, the overall positivity for COX-1 and COX-2 for the patient groups was 33% and 28%, respectively. However, there was no direct correlation between COX-1 expression and COX-2 expression for an individual patient.
To determine the significance of COX-2 overexpression, all patients were divided arbitrarily into two groups according to their COX-2 expression status: 1) the COX-2 negative group (n = 54 patients) and 2) the COX-2 positive group (n = 21 patients). COX-1 overexpression was neglected in the analysis due to its lack of prognostic significance. The clinical profiles of the COX-2 negative group and the COX-2 positive group are listed in Table 1. There were no correlations between COX-2 overexpression and the age or performance status of the patients. A further attempt was made to determine a correlation between COX-2 expression and several clinical parameters. No significant difference was found in the histopathologic subtypes or in the tumor shape between the two groups. Neither the tumor size nor the extent of parametrial involvement was correlated with COX-2 expression. Even though patients in the COX-2 positive group showed a tendency to have a higher incidence of pelvic lymph node metastases, the difference was not statistically significant.
Table 1. Clinical Profiles
No. of patients (%)
COX: cyclooxygenase; LCK: large cell keratinizing; LCNK: large cell nonkeratinizing; LN: lymph node; CT: chemotherapy; RT: radiotherapy; NS: not significant; Gy: grays.
Although the initial treatment response for patients in both groups was generally good, there was a trend toward a delayed regression after the initial treatment for patients in the COX-2 positive group compared with patients in the COX-2 negative group. There were a total of 20 recurrences among all patients at the time of this analysis. Eleven of 54 patients (20%) in the COX-2 negative group had suffered recurrences, whereas 9 of 21 patients (43%) in the COX-2 positive group had experienced recurrences. This higher incidence of treatment failure for patients in the COX-2 positive group was statistically significant (P < 0.05). Figure 2 illustrates the patterns of treatment failure for both groups. Three recurrences were central failures, four recurrences were lymph node failures, and four recurrences were hematogenous metastases in the COX-2 negative group, whereas six patients had central failures, eight patients had lymph node failures, and one patient had a hematogenous metastasis in the COX-2 positive group. Despite a similar rate of complete remissions with the initial treatment, the central failure rate among patients in the COX-2 positive group was much higher compared with the rate among patients in the COX-2 negative group (29% vs. 6%; P = 0.012). The overall lymph node recurrence rates, including pelvic, para-aortic, and/or supraclavicular lymph node failure, differed significantly between the two groups (38% of patients in the COX-2 positive group vs. 7% of patients in the COX-2 negative group; P = 0.003). In contrast, the rate of hematogenous metastases was not correlated with the COX-2 status. The overall incidence of hematogenous metastases for patients in the COX-2 positive and COX-2 negative groups was 5% and 7%, respectively. It is interesting to note that COX-2 overexpression in tumor cells was correlated with a shorter interval to tumor recurrence, and a remarkable difference in disease free survival was found between patients in the two groups. The median interval to disease recurrence of 26 months for patients in the COX-2 negative group was comparable to the median interval of 9 months for patients in the COX-2 positive group (P = 0.03; log-rank test) (Fig. 3).
Survival and Prognostic Factors
With a minimum follow-up of 60 months, the 5-year overall and disease free survival rates for all patients who were treated with radiotherapy and concurrent chemotherapy were 85% and 65%, respectively. When patients were divided into two groups according to positive or negative COX-2 expression status, patients in the COX-2 positive group had a poorer prognosis compared with patients in the COX-2 negative group. The overall 5-year actuarial survival rate was 56% (95%CI, 33–79%) for patients in the COX-2 positive group and 94% (95%CI, 87–99%) for patients in the COX-2 negative group (P = 0.0003; log-rank test). The 5-year disease free survival rate was 47% (95%CI, 25–69%) for patients in the COX-2 positive group and 81% (95%CI, 70–92%) for patients in the COX-2 negative group, and the difference was statistically significant (P = 0.003; log-rank test). The overall survival curves and disease free survival curves for the two groups are shown in Figure 4. Because the prognosis of the patients was correlated with several clinicopathologic variables, an attempt was made to determine the influence of COX-2 expression on patient survival by using Cox regression analyses. In the initial univariate analysis for the overall actuarial and disease free survival rates, the extent of parametrial involvement and COX-2 overexpression were important prognostic factors (Table 2), but only COX-2 overexpression remained an independent prognostic factor in the multivariate analysis (Table 3).
Table 2. Univariate Analysis of the Prognostic Factors
The current study showed that COX-2 overexpression was detected immunohistochemically in approximately 30% of patients with FIGO Stage IIB SCC of the uterine cervix. In addition, patients in the COX-2 positive group had a higher likelihood of central and lymph node failure and a shorter interval to tumor recurrence compared with patients in the COX-2 negative group. This suggests that COX-2 overexpression is responsible for the more invasive phenotype of tumor cells, reflecting the biologic aggressiveness of SCC in the uterine cervix. Moreover, COX-2 overexpression was correlated directly with a particularly poor prognosis and was an independent prognostic factor in both univariate analysis and multivariate analysis. Although tumor volume, lymph node metastasis, and FIGO stage all have been recognized as important forecasters for patients' survival, the results of this study indicate that COX-2 overexpression is also a potent molecular marker of tumor recurrence and poor prognosis with clinical significance that surpasses the other well-known clinicopathologic parameters in patients with FIGO Stage IIB SCC of the uterine cervix who undergo radiation and concurrent chemotherapy.
It is well known that the concentration of prostaglandin in tumor tissues is much greater than its concentration in corresponding normal tissues.13–15 Biochemical evidence of enhanced biosynthesis of prostaglandin by the COX-1 and COX-2 enzymes has been supported by numerous experimental models.14, 15 Two isozymes are found predominantly in the same organelles, at the endoplasmic reticulum and nuclear envelope of cells.28 Although both enzymes have very similar structures, catalytic mechanisms, products, and kinetic properties, there are distinct differences in the regulation of gene expression and their metabolic or physiologic functions.13 Increased levels of COX-1, which is a housekeeping gene, are responsible for normal physiologic production of prostaglandin that is required for gastrointestinal, renal, and vascular homeostasis, whereas alterations in COX-2 expression are observed more frequently in several types of malignant tumors, because COX-2 is present independently in cells only during the early stages of cell differentiation or replication.13, 29 Nevertheless, large variations in COX-1 levels occasionally exist in some types of tumors.13, 17 Doré et al. reported that the expression of COX-1 protein, but not COX-2 protein, was substantially higher in ovarian adenocarcinoma.30 Hwang et al. also showed that higher levels of the COX-1 protein were detected in breast carcinoma compared with in benign tumors or other normal tissue specimens.31 Although COX-1 expression in uterine cervical carcinoma is not understood fully, the immunohistochemical study in this investigation demonstrated that COX-1 staining was observed only in the adjacent normal tissues and not in the tumor cells, whereas COX-2 staining was observed in epithelial tumor cells. Moreover, increased COX-1 expression was neither a determinant of COX-2 expression nor an independent prognostic factor in patients with carcinoma of the uterine cervix.
In previous studies of the clinical significance of COX-2 expression in patients with uterine cervical carcinoma, Ryu et al. reported that elevated COX-2 expression was associated with a greater incidence of parametrial invasion and lymph node metastases in patients with FIGO Stage IB carcinoma of the uterine cervix who underwent surgery.11 Gaffney et al. also demonstrated that, irrespective of disease stage, there was a positive correlation between COX-2 expression levels and prognosis in patients with FIGO Stage IB–IIIB who were treated with radiotherapy alone.12 Our observation also identified a subpopulation with COX-2 overexpression among patients with FIGO Stage IIB disease who underwent radiation and concurrent chemotherapy as a subset of patients with a poor prognosis. Even if their tumors were not particularly bulky or if they had other features that were more suggestive of an unfavorable prognostic factor, the patients with COX-2 overexpression had worse overall actuarial survival and disease free survival rates with a greater incidence of central and lymph node failure. This is consistent with patient cohorts who underwent either surgery or radiotherapy. It is interesting to note that patients in the COX-2 positive group were associated with a delayed response and a rapid recurrence compared with patients in the COX-2 negative group, suggesting that the COX-2 gene may play a role in the cell cycle response and the apoptotic response to chemotherapy and radiotherapy.
The pathophysiologic process involved by COX-2 in uterine cervical carcinoma requires further investigation in tumor cell lines and clinical research to better understand their interaction. However, it appears that COX-2 may play a very important role in modulating the local invasiveness and metastatic potential of uterine cervical carcinoma.20, 23 The increased invasiveness of carcinoma is associated with both the activation of type 2 membrane metalloproteinase (MMP-2) and the increased quantity of mRNA for MMP-1.23, 29, 32 These enzymes, which are induced by COX-2, digest the collagen matrix of the basement membrane, stimulating the invasive and motile phenotype of the tumor cells.29 In addition, administering a nonsteroid anti-inflammatory drug (NSAID) significantly decreases cell growth and increases apoptosis, which is conferred by the increased COX-2 levels.21, 22, 29 Apart from the effects on cellular adhesion and apoptosis, COX-2 overexpression has been implicated as a possible modulator of tumor angiogenesis.10, 24, 29 The pharmacologic inhibition of COX-2 in vitro has led to the decreased production of proangiogenetic cytokine and of vascular endothelial growth factor (VEGF), which may be related to the reduced tumor growth in COX-2 null mice.12, 29, 33 Moreover, a combination of neutralizing anti-VEGF antibodies and radiation further enhanced the efficacy of the cell-killing effect of the ionizing radiation in murine tumor model systems.33 Taken together, the data concerning the biologic effects of the COX-2 enzyme affecting many pathophysiologic processes makes it suitable as an attractive molecular target for novel therapeutic intervention.
Identifying selective COX-2 inhibitors as a therapeutic approach would provide new insights into the treatment of patients with malignant disease. Compared with COX-1 inhibitors, which are accompanied by a high incidence of gastrointestinal side effects, it is noteworthy that selective COX-2 inhibitors cause much less damage to the mucosa of the gastrointestinal tract than classic NSAIDs.18, 24, 29 Additionally, several reports have demonstrated that selective COX-2 inhibitors have potential as chemosensitizers or radiosensitizers for treating human malignancies.34–37 Duffey et al. and others have shown that NSAIDs significantly increased the cytotoxicity of various chemotherapeutic agents in different cell lines, including human lung carcinoma cells and leukemia cells.35 Furthermore, treatment with a COX-2 inhibitor not only greatly enhanced the tumor response without markedly affecting the normal tissue radioresponsiveness but also further increased the intrinsic tumor cell radiosensitivity in animal tumor models.36, 37 This suggests that selective COX-2 inhibitors would be beneficial if administered in combination with either chemotherapy or radiotherapy. However, future clinical trials are needed to determine whether the efficacy of chemotherapy or radiotherapy can be enhanced by cotreatment with a selective COX-2 inhibitor.
In conclusion, COX-2 overexpression can be used to predict the prognosis of patients with FIGO Stage IIB SCC of the uterine cervix who are treated with radiotherapy and concurrent chemotherapy. Because immunohistochemical analysis is a relatively simple technique that is feasible in daily clinical practice, COX-2 expression should be an important, routine part of the management of such patients. Further study is recommended to determine the clinical usefulness of selective COX-2 inhibitors in treating patients with uterine cervical carcinoma who have COX-2 overexpression.