Our results demonstrate that the COX-2 enzyme is universally expressed in adenocarcinoma of oesophagus but the amount of COX-2 expression is variable. Tumours expressing high levels of COX-2 are associated with poor patient survival after surgery. We also demonstrated a relationship between COX-2 expression and higher T and N stage as well as increased risk of tumour recurrence after curative resection. On further analysis we identified a cut off score of ≥200 for COX-2 expression, which could serve as a marker of aggressive disease and poor survival. We demonstrated a positive correlation between COX-2 m-RNA expression detected by PCR and COX-2 protein expression detected by immunohistochemistry in patients with oesophageal adenocarcinoma. This further strengthens the validity of our immuno-scoring system and shows a link between genomics and proteomics.
We evaluated COX-2 expression by immunohistochemistry as this is one of the most widely used method that can be used to analyze protein expression in paraffin embedded archival material. Using immunohistochemistry, the signal can be precisely localized to the tumour cell which is not possible when using other techniques like immuno-blot and PCR which can overestimate COX-2 expression by identification of expression within both inflammatory and neoplastic cells. One of the strengths of our work is the specificity of our antibody, as demonstrated by the absence of staining in the antigen adsorption study and in the negative controls. Quantification of protein expression may be difficult when using immunohistochemistry. However, we used an intensity-proportion scoring system which has been extensively used in breast cancer scoring. We found that this scoring system was very easy to use and highly reproducible. The strength of our scoring system was further confirmed by our PCR data, which showed a positive correlation between RT-PCR quantification and immuno-scoring for COX-2.
The literature is fairly divided on the clinicopathologic significance of COX-2 in oesophageal cancer. There are two studies addressing this issue in patients of Japanese origin with squamous cell cancer of oesophagus and have demonstrated a lack of correlation between COX-2 expression and clinicopathologic features of the tumour or overall survival [24, 25]. Two additional studies have been performed in patients of western origin with adenocarcinoma of oesophagus and like our study have demonstrated a negative correlation between levels of COX-2 expression and overall survival, although unlike our study no correlation was found between COX-2 expression and clinicopathologic features of the tumour and no attempt was made to correlate the COX-2 protein expression with the COX-2 m-RNA expression[23, 26]. Our study substantiates the findings of these previous studies and gives a better insight into the link between the clinico-pathologic features and COX-2 expression in adenocarcinoma of oesophagus. However, another previous study by Lagorce et al  found that COX-2 is expressed predominantly in well differentiated adenocarcinoma and did not find any correlation between COX-2 expression and patient survival or any of the clinicopathologic features of the tumour. It is difficult to explain the differences in the findings of this study as compared to ours and the other two studies. However, the study by Lagorce et al included patients with distant metastasis and grouped moderate and well differentiated tumours together producing a slightly skewed distribution in favour of poor differentiation (47 vs 19). Our study excluded patients with distant metastasis and grouped moderate and poorly differentiated tumours together to produce a skew in favour of well differentiation (79 vs 11). These differences in patient selection and study design might explain the differences in results.
The strengths of our study include the large number of enrolled consecutive patients with adenocarcinoma of oesophagus, the use of immunohistochemistry to precisely localise the expression in the neoplastic cells, the use of resection specimens rather then superficial biopsies (as biopsies could underestimate the expression of COX-2 due to its heterogenous nature of expression), the demonstration of a positive link between COX-2 protein (immunohistochemistry) and m-RNA (PCR) expression and the specificity of our antibody as shown by the absence of staining in negative controls and during antigen adsorption study.
Given the current literature and the findings of our study it can be concluded that the prognostic significance of COX-2 overexpression in oesophageal cancer varies with the histological subtype of the tumour, a negative prognostic influence in adenocarcinoma but with no prognostic influence in squamous cell cancer. This raises the possibility that the biological behaviour of the oesophageal cancer might vary with the histological type of the tumour.
It is not clear from our study whether COX-2 plays a primary role in carcinogenesis and progression or is simply a para-phenomenon. However, a primary role of COX-2 is consistent with the epidemiological data linking increased survival and reduced incidence of oesophageal cancer and NSAID use. The precise mechanism for the survival disadvantage observed in patients with high levels of COX-2 is not clear but it is well established that cancer cells expressing high levels of COX-2 have a higher rate of proliferation and a lower rate of apoptosis. Selective inhibition of COX-2 reduces tumourigenesis in different models of carcinogenesis[17, 19, 30]. Thus, COX-2 has been shown to play a causal role in carcinogenesis. Furthermore, COX-2 expression has been linked to enhanced levels of angiogenesis and expression of metalloproteinases resulting in an increased invasiveness and metastatic potential of tumours. This is further demonstrated by the effect of COX-2 inhibitor in suppressing the haematogenous metastasis of colon cancer in mice. Our study demonstrates an association between high levels of COX-2 expression and higher T-stage, N-stage and an increased risk of tumour recurrence after potentially curative resection. These are all features of a tumour with increased malignant and metastatic potential and this might be one of the mechanisms explaining the link between poor survival and increased COX-2 expression. However, on multivariate analysis we found that the effect of COX-2 on survival was independent of T and N stage. This raises the possibility that there may be other mechanisms by which COX-2 adversely influences survival.
The adverse influence of COX-2 on patient survival has also been reported in other gastrointestinal tumours like the stomach and colon but none of these studies elucidate the mechanisms behind the influence of COX-2 on survival, which still remains a matter for future research.