The COX enzyme is probably the most common therapeutic drug target in human history. Aspirin, a COX Inhibitor, has been used for almost 4000 years, and large amounts of these compounds are consumed each year. Research in this area has been dominated by investigations into the COX enzymes, also known as prostaglandin-endoperoxide synthases, which are central and rate-limiting enzymes in the biosynthesis of prostaglandins (Arber 2008; Tuynman 2004). Three COX isoforms have been identified: COX-1, COX-2, and COX-3.

COX-1 and COX-2 are located on different chromosomes and their expression is tightly regulated (Tuynman 2004). COX-1 is mapped to chromosome 9q32-q33.2, is encoded by the PTGS1 gene, and constitutively expressed in normal tissues. It serves as a ‘housekeeper’ of mucosal integrity. COX-1 is the central enzyme in the biosynthetic pathway to prostaglandins from arachidonic acid, it produces prostacyclins, prostaglandins, and thromboxane, which protect gastric mucosa and play a key role in platelet aggregation and renal microvasculature dynamics. COX-2 is mapped to chromosome 1q25.2-q25.3, and is encoded by the PTGS2 gene, an immediate early response gene that is highly inducible by either neoplastic or inflammatory stimuli. COX-2 is involved in the synthesis of prostaglandins and thromboxanes, which are regulators of processes that are relevant to cancer development. It is generally accepted that alterations in COX-2 expression and the abundance of its enzymatic product prostaglandin E₂ (PGE₂) have key roles in influencing the development of CRC.

COX-3, a third distinct COX isozyme is a COX-1 variant formed by intron retention, a form of alternative splicing (Chandrasekharan 2002). COX-3 shares all the catalytic features of COX-1 and -2; however, its exact role is yet to be fully understood (Chandrasekharan 2002).

Relative to normal mucosa, COX-2 overexpression occurs in about half of CR adenomas and in 85% of human CRCs, making COX-2 an attractive therapeutic target (Elder et al. 2002; Sheehan et al. 1999). Moreover, the fact that COX-2 expression is up-regulated in both pre-malignant and malignant CR tissue has also potential implications for the prevention of this type of cancer. Already 40 years ago, NSAIDs were hypothesized to inhibit the growth of CRC after a significant decrease in PGE₂ was observed in CRC tissue compared to the normal surrounding mucosa (Bennett and Del Tacca 1975; Jaffe 1974). The preventive efficacy of this class of agents is supported by more than 300 animal studies. Most significantly, 70 out of 72 epidemiological studies clearly demonstrated that NSAID/aspirin consumption prevents adenoma formation and decreases the incidence of, and mortality from, CRC (Fig. 1). However, NSAID consumption is not free of toxicities. There are well-known serious adverse events to the gastrointestinal, renal, and cardiovascular systems. In the United States alone, 260,000 hospitalizations and 26,000 deaths were attributed to NSAID consumption in 2002 (Grover et al. 2003).

Since COX-2-selective inhibitors do not inhibit COX-1, they are not generally believed to harm the normal mucosa. However, because COX-2 is overexpressed throughout the multistep process of CRC carcinogenesis, they would seem to be an ideal drug candidate for use in the healthy population for the prevention of CRC. In the early 1990’s, pharmaceutical companies began developing COX-2 selective inhibitors with minimal effect on COX-1 activity (Arber 2008). In 1999 and 2000 three international, multicenter, prospective, randomized, placebo-controlled trials in the secondary prevention of CRC were launched (Baron et al. 2006; Bertagnolli et al. 2006, 2009; Bresalier et al. 2005). These clinical trials demonstrated the efficacy of COX-2 inhibitors as a strategy for reducing cancer incidence, although associated side effects and in particular cardiovascular (CVS) side effects prevented their routine use in the general population.

In the Prevention of Colorectal Sporadic Adenomatous Polyps (PreSAP) study, 1,561 patients from 107 sites in 32 countries were recruited. Celecoxib reduced adenoma recurrence by a third after one and three years (p < 0.001). Celecoxib was particularly potent in inhibiting the recurrence of advanced adenoma by 51 % (Arber et al. 2006). The Adenoma Prevention with Celecoxib (APC) trial enrolled 2,035 patients that were randomized to receive placebo, celecoxib 200 or 400 mg bid. In patients taking celecoxib, polyp recurrence was reduced by 33 and 45 % for patients taking 400 and 800 mg of the drug, respectively (p < 0.0001). The relative risk of advanced adenomas was even more drastically reduced: by 57 and 66 %, respectively (p < 0.0001) (Bertagnolli et al. 2006). It was shown that compared to placebo, patients taking celecoxib had fewer and smaller adenomas as well as reduction in overall tumor burden. In a third study the Adenomatous Polyp Prevention on Vioxx (APPROVe), 2,547 participants were randomized to receive rofecoxib at 25 mg qd or placebo. A 25 % reduction in polyp recurrence was seen after one and three years, the effect on advanced adenoma was almost identical (RR-0.76 (95 % CI 0.69–0.83)) (Lagaos 2006).

However, all three studies were terminated earlier than planned due to substantial concern of increased cardiovascular system (CVS) toxicity, as seen by an increase in cardiovascular events (Bertagnolli et al. 2009; Bresalier et al. 2005). The CVS toxicity seen in the APPROVe trial prompted Merck to withdraw rofecoxib from the market; this decision was made even before the efficacy of the drug was evaluated. In the APC trial, the CVS toxicity, as evaluated by an independent cardiovascular adjudicating committee, increased from 1.0 % (n = 7/679) for placebo to 2.5 % (n = 16/685), and 3.4 % for celecoxib (200 and 400 mg bid, respectively) (p < 0.01). As a result, the NCI to suspended the trial. Lastly, the proportion of all patients experiencing CVS toxicity in the PreSAP trial increased from 1.9 % (n = 12/628) for placebo to 2.5 % (n = 23/933) for celecoxib (400 mg qd) (p = NS).

The CVS toxicity persisted 1 year after rofecoxib was discontinued (APPROVe) (Lagaos 2006) and 2 years after celecoxib was discontinued (PreSAP and APC) (Bertagnolli et al. 2006; Arber et al. 2006) trials. Of note is the disparity in CVS toxicity from celecoxib between the APC and PreSAP trials (Arber 2008). A plausible explanation for this discrepancy is the difference in dosages. The APC trial gave celecoxib twice daily, for a total daily dose of 400 or 800 mg. It stands to reason that a greater dose increases the likelihood of an adverse reaction. Another plausible explanation for the discrepancy is that the 400 mg given once daily in the PreSAP trial was less toxic than the 200 mg given twice daily in the APC trial because of the relatively short half-life of celecoxib.

The actual extent of the CVS risk associated with COX-2 selective inhibitors remains unclear (Arber 2008). The trials were not designed to assess for cardiovascular events and it was difficult to control for confounding variables. Most importantly, the number of events was very low, and the vast majority of patients tolerated celecoxib without the related toxicity throughout the study (Bertagnolli 2007). The polyp recurrence rate reduction was the same after one and three years in all three studies. Cardiovascular toxicity started to increase only after 12–18 months. This suggests the possibility that use of COX-2 inhibitors for 1 year may be sufficient to prevent polyp recurrence, before toxicity appears. The gastrointestinal toxicity of celecoxib in the PreSAP and APC trials has also been recently adjudicated (Arber et al. 2011). There was no significant difference between the drug and placebo for the entire 3 year duration of the study. The discovery of CVS toxicity related to COX-2 specific inhibitors has made the development of new agents in this field difficult. However, to ignore potential benefit from chemoprevention is to accept a higher than necessary death rate from CRC.

The exact mechanism by which COX-2 inhibitors exert their anticancer properties is currently unknown. As mentioned above, the involvement of COX-2 in CR tumorigenesis has been attributed to its role in the production of PGE₂ which its levels were found elevated in CR cancers. Thus, deregulation of the COX-2/PGE₂ pathway appears to affect CR tumorigenesis via a number of distinct mechanisms involving promotion of tumor maintenance and progression, induction of metastatic spread, and others (Greenhough et al. 2009). There are at least seven mechanisms underlying the pro-tumorigenic effects of COX-2; (Tuynman 2004):

COX-2 inhibitors can also act through COX-2-independent pathways. They can induce apoptosis in cancer cells not expressing the COX-2 enzyme. A variety of non-COX-2 targets for COX-2 inhibitors have been suggested, such as, NF-kB, peroxisome proliferator activating receptor-δ and -γ, protein kinase G, and Bcl-XL (Grover et al. 2003; Rao and Reddy 2004; Sinicrope 2006; Arber and Levin 2008).

Personalized medicine has remained an elusive goal and its utilization in chemoprevention is greatly anticipated. If COX-2 inhibition is the principal mechanism through which NSAIDs work, then these agents should be targeted at tumors that overexpress COX-2. Previous studies have shown that aspirin reduces the risk of CRC in COX-2 expressing cancers, but is not effective in COX-2 negative cancers (Chan et al. 2007). The efficacy and toxicity of COX-2 inhibitors may be affected by polymorphisms in COX-2, COX-2 targets, and related metabolizing enzymes (Arber 2008; Ulrich and Bigler 2006). It was suggested that polymorphisms in, COX-2 itself and metabolizing enzymes such as, uridine diphosphatidyl glucotransferase, may increase chemopreventive efficacy by up to 50 % (Macarthur et al. 2005; Lin et al. 2002). Moreover, polymorphisms in COX-2, and particularly -1195A > G may modulate the genetic susceptibility for CRC onset in some cases (Pereira et al. 2010). Another COX-2 polymorphism (rs4648319) was found to modify the effect of aspirin, supporting a role for COX-2 in the etiology of CRC and as a possible target for aspirin chemoprevention (Barry et al. 2009). It appears that polymorphisms in COX-2 targets or metabolizing enzymes may affect COX-2 efficacy and/or toxicity. However, the current literature on these interactions is still very limited (e.g., COX1 P17L or COX2 -765G > C). Reliable detection of gene-COX-2 interactions will require greater sample sizes, consistent definitions of COX-2 use, and evaluation of the outcome of chemoprevention studies. Nevertheless, these studies suggest that this genetically based higher-risk group definition may help to shift the balance between risk and benefits for the use of COX-2 inhibitors in chemoprevention that is currently hampered by adverse side effects (Pereira et al. 2010).