Introduction

Colorectal cancer (CRC) represents a major medical challenge. Despite significant progress, the mortality of advanced CRC has remained unchanged for decades [1]. Although screening colonoscopy can prevent CRC, it is not widely employed because of poor patient acceptance and cost [2]. Virtual colonoscopy has its own limitations, and its application to CRC screening is even more limited [3]. These considerations provide renewed impetus for the development of CRC chemoprevention. Nonsteroidal anti-inflammatory drugs (NSAIDs) are currently the most promising class of compounds for the prevention of CRC. Here, we review salient points of their efficacy and safety and assess their prospects as the agents of choice for the chemoprevention of CRC.

NSAIDs Old and New

NSAIDs have been in development for over 100 years, beginning with the synthesis of acetylsalicylic acid or aspirin in 1897 [4]. This seminal event was the culmination of over two centuries of efforts beginning with the 1763 letter by the Reverend Edward Stone to the Earl of Macclesfield describing the use of powder derived from willow bark for treating fever in 50 patients. Salicin, the Latin word for willow, served as the basis for the term salicylic and its derivatives.

Conventional NSAIDs are classified based on their structural features or on their ability to inhibit cyclooxygenase (COX), with their selectivity for COX-2 inhibition considered a defining characteristic, e.g., COX-2 selective inhibitors [5]. The structural similarities between them are few, with the exception of a carboxylic moiety that is present in nearly all of them. There are currently at least 20 approved NSAIDs in the USA and more elsewhere.

Seemingly incessant efforts to generate improved NSAIDs have led to several new (non-FDA approved yet) compounds, all chemically modified NSAIDs. Nitro-NSAIDs are NSAID derivatives in which a nitric oxide releasing moiety (−ONO2) is attached covalently to NSAIDs via an aromatic or aliphatic linker [6]. Several novel sulindac derivatives have been generated [7] including sulindac benzylamine [8, 9] and des-methyl-sulindac sulfide and its pro-drug des-methyl-sulindac that fail to inhibit COX-2 activity [10]. Esterification/amidation of the carboxylic acid moiety in various NSAIDs, e.g., indomethacin [11], has been pursued to obtain selective COX-2 inhibitors [12]. Phosphatidylcholine (PC) NSAIDs have been synthesized to remove the gastrointestinal (GI) toxicity of conventional NSAIDs by capitalizing on the ability of PC to protect the gastrointestinal mucosa without diminishing their pharmacological efficacy [13]. Finally, phospho-NSAIDs are derived by a fairly simple chemical modification of existing NSAIDs. In particular, their carboxylic moiety is often esterified to a dialcohol spacer moiety; the second –OH of the spacer is linked to a diethylphosphate moiety, which enhances cellular uptake [14]. Recently, one of the phospho-NSAIDs was pegylated to improve its pharmacokinetics and bioavailability [15].

Chemopreventive Efficacy

That NSAIDs can prevent CRC is now well established thanks to an extensive and fairly consistent body of work by many investigators around the world. Such work includes huge epidemiologic studies with hundreds of thousands of subjects in the aggregate. The clear conclusion is that the prolonged use of NSAIDs reduces adenomatous polyps, incident disease, and death from CRC by 30–50 % [16–18]. Most of the interventional studies have focused on aspirin and sulindac.

Studies with Aspirin and its Derivatives

Randomized clinical trials (RCTs) using the development of colon adenomas as surrogate endpoints for CRC prevention demonstrated the ability of aspirin to prevent CRC. The results of three such trials were congruent, but the effect was modest and somewhat disappointing, as the risk of recurrent adenoma was reduced at most by one third [19–21].

Rothwell et al. [22] examined the cancer outcomes of four randomized trials originally designed to evaluate the effect of aspirin 75–1200 mg/day on the prevention of cardiovascular events in populations including subjects at low and high risk. Treatment with aspirin 75–500 mg/day reduced the 20-year risk of CRC by 24 % and the CRC-associated mortality by 35 %. The same group examined data from eight cardiovascular prevention RCTs of daily aspirin included in a study that pooled individual patient data and examined the effects of randomized aspirin treatment on all cancer mortality [23••]. Aspirin 75–1200 mg/day was associated with 21 % lower risk of death from any cancer. What was rather surprising was the 5-year lag until the benefit became evident. This lag also applied to the risk of death due to CRC (HR = 0.41; p = 0.05), whose reduction began 5 years after the initiation of aspirin treatment. Analysis of 51 RCTs of daily low-dose aspirin for primary prevention revealed that aspirin reduced cancer incidence beginning on the third year of administration [24••]. Although these data are compelling, the reservation has been raised that these trials were not originally designed to examine CRC incidence or mortality [17, 25].

Sosters et al. [25] have recently argued that the antiplatelet action of aspirin plays a central role in its antitumor effect, since daily low-dose aspirin given for the prevention of cardiovascular disease events prevents adenoma recurrence and decreases the incidence of CRC and attributable mortality. We have speculated similarly in the past, based on the profile of eicosanoids in human colon cancers [26].

NO-aspirin administered for CRC prevention was both efficacious and safe, at least in terms of gastric toxicity [6, 27]. All three positional isomers of NO-aspirin showed promise as anticancer agents [28], but the clinical evaluation of their role in CRC prevention was not pursued because of concerns regarding potential genotoxicity.

Both aspirin-PC and ibuprofen-PC showed in vitro and in vivo properties suggesting their potential utility in patients at risk for CRC [29]. For example, both inhibited the growth of colon cancer cells in vitro and the development of colonic aberrant crypt foci (ACF) in the azoxymethane rat model of CRC. Clinical studies with these compounds are apparently in progress.

Studies with Sulindac and Related Compounds

In a well-reasoned effort to overcome the limited chemopreventive efficacy of conventional NSAIDs, Gerner and Meyskins have developed a combination approach [30]. Their biochemical goal was to reduce the intracellular levels of polyamines, which control, at least in part, colonocyte proliferation. To this end, they used a combination of difluoromethylornithine (DFMO) and sulindac. DFMO is a suicide inhibitor of ornithine decarboxylase, which catalyzes the rate-limiting step in polyamine synthesis. Sulindac, on the other hand, increases cellular export of polyamines by activating the spermidine/spermine acetyltransferase. This combination produced impressive results both in preclinical studies and in a phase III trial in patients with prior colon polyps, reducing total metachronous colorectal adenomas by 70 % and advanced and/or multiple adenomas by >90 % [31]. This regimen was well tolerated; in particular, there were no auditory side effects induced by DFMO. The combination of aspirin and folate produced no benefit from folate [32].

Takayama and Niitsu have for years pursued assiduously as chemoprevention targeting ACF, aggregates of colonic crypts, considered precursors of polyps. Their reasoning was that the low complexity of ACF would make easier their elimination by chemopreventive agents [33–35]. A 12-month interim analysis of a RCT of sulindac 300 mg or etodolac 400 mg given daily for the first two months showed that sulindac was able to reduce aberrant crypt foci by 41 % at 2 months and the number of polyps by 51 % at 12 months [36••]. Of note, the COX-2-specific inhibitor etodolac showed no such effect. They also observed that this effect of sulindac was similar in magnitude to that obtained after long-term administration of NSAIDs with polyp recurrence as the endpoint. Responsiveness of ACF to the chemopreventive agent may predict its effect on polyps. Thus, short-term or even discontinuous administration of NSAIDs such as sulindac may prevent CRC. The newer, more efficacious derivatives of sulindac, e.g., phospho-sulindac [14], may have a role in such an approach.

Sulindac benzylamine, a novel sulindac derivative, does not inhibit COX-1 or COX-2, yet potently inhibits the growth and induces the apoptosis of human colon tumor cells [9]. The basis for this activity appears to involve cyclic guanosine 3′,5′,-monophosphate phosphodiesterase (PDE); the PDE5 isoform is essential for colon tumor cell growth. Other sulindac derivatives such as des-methyl-sulindac sulfide and its pro-drug des-methyl-sulindac failed to significantly inhibit tumor formation in APC/Min mice [10], a result attributed to limited activation to their active forms in vivo.

The last group of chemically modified NSAIDs that has been studied in the prevention of CRC is that of phospho-NSAIDs, with phospho-sulindac (PS) being the one most extensively studied. PS is >10-fold more potent than sulindac in inhibiting the growth of cultured CRC cells and more efficacious in the prevention of CRC [14, 37••]. Like sulindac, PS synergized with DFMO to prevent CRC; in Apc/Min mice, the PS/DFMO combination reduced tumor multiplicity by 90 %. The mechanism of action of PS/DFMO appears complex. Two prominent effects are the induction of oxidative stress and the suppression of polyamine levels. A very interesting aspect of the effect of PS/DFMO is the strong suppression of the thioredoxin system [37••], an emerging regulator of chemoprevention [38].

PS was studied extensively for its safety, a critical parameter for any candidate chemopreventive agent. Multiple studies have shown its safety to be equivalent to that of placebo and far superior to that of sulindac [14]. Although not fully understood, the safety of PS is partly attributed to its unique pharmacokinetic properties: the blood AUC0–24 h of PS is around 40 % of that produced by an equimolar amount of sulindac, while in the stomach, the main site of NSAID toxicity, PS is mainly intact and its two gastrotoxic metabolites, sulindac and sulindac sulfide, are present in miniscule amounts [39]. PS’s more rapid detoxification by cytochrome P450s and flavin monooxygenases seems to contribute to its safety [40].

Genetic Syndromes: Familial Adenomatous Polyposis and Lynch Syndrome

The chemoprevention efficacy of NSAIDs has been evaluated in familial adenomatous polyposis (FAP) and Lynch syndrome (hereditary non-polyposis colorectal cancer). These syndromes not only constitute a pressing clinical need but also provide the opportunity to conduct expedited studies on CRC chemoprevention. Sulindac and the COX-2 selective inhibitors celecoxib and rofecoxib reduced the size and number of colorectal polyps after 6–9 months of treatment in FAP patients [41–43]. Aspirin 600 mg/day, however, failed to significantly reduce the number of polyps in the sigmoid colon and rectum [44••]. Similarly, aspirin 100 mg/day showed limited efficacy in another trial that was largely inconclusive due to small subgroup size [45]. In Lynch syndrome, aspirin 600 mg/day for up to 4 years was effective in preventing CRC [46••]. For those taking aspirin for 2 years or longer, the hazard ratio was 0.41 (p = 0.02). This trial was the first RCT of aspirin with CRC as the primary endpoint. The evidence is strong that aspirin is an effective chemopreventive agent in hereditary cancer with an effect equivalent to that achieved with surveillance colonoscopy.

Mechanistic Aspects of NSAIDs

The mechanistic effects of NSAIDs in cancer have been conceptualized as COX dependent and COX independent. Initially, the anticancer activity of NSAIDs was considered to derive solely from its COX inhibitory effect [47]. This speculation was based primarily on the well-recognized ability of aspirin and NSAIDs to inhibit COX, the enzyme catalyzing the synthesis of prostaglandins, and on the increased PGE2 levels in CRC [26]; PGE2 stimulates the proliferation of CRC cells [48], an effect blocked by NSAIDs [49]. However, additional work pointed out that COX-independent effects may also contribute greatly to the chemopreventive effect of NSAIDs [50]. Multiple signaling pathways have been implicated, both COX dependent and COX independent. Key players seem to be NF-κB, the Wnt pathway, the DNA mismatch repair system, NAG-1 (NSAID-induced gene) [51], and sphingosine-1 kinase [52]. Induction of oxidative stress may play a role, at least where it is the initial event that activates redox-responsive signaling cascades culminating in apoptotic cell death [38, 53]. The modified NSAIDs seem to work both via COX-dependent and COX-independent pathways, but redox effects are prominent [54••]. Recent work is exploring whether epigenetic regulation by NSAIDs may contribute to their chemopreventive effects [55, 56]. Despite decades of work, no unifying mechanism has been obtained [47, 57, 58], indicating either that different mechanisms may operate in particular instances [59] or that the actual mechanism awaits discovery.

The Safety of NSAIDs

Safety, important for any medication, is critical in chemoprevention as the NSAID will likely be taken for years when its side effects may be compounded. Of note, the side effects of NSAIDs are broad and can be lethal [18, 60]. The most consequential of them are GI and renal toxicity and, to a lesser extent, cardiovascular, CNS, and platelet side effects. There is no safe NSAID.

Aspirin, the oldest and most extensively used, has the greatest safety database of all NSAIDs. A close parallel to the chemoprevention use of aspirin is its long-term administration for the prevention of cardiovascular events, a standard clinical practice for many years. The GI side effects of aspirin have been the subject of numerous studies [61]. Compared to those who did not use it, aspirin not only increased by 1.6–3.1 times the RR for GI bleeding but also increased the risk for adverse GI symptoms (OR = 1.7), such as nausea and dyspepsia [62]. The GI toxicity of aspirin increased with increasing dose and with advancing patient age. For example, 2.5 % of those taking more than 100 mg of aspirin per day had GI bleeding compared with 1.1 % of those taking fewer than 100 mg/day [63].

The cardiovascular complications of NSAIDs and COX-2 selective inhibitors are a topic of major concern and significant complexity [64–66]. Suffice it to mention here that most nonselective NSAIDs and COX-2 selective NSAIDs (coxibs) increase the risk of adverse cardiovascular events, including death, myocardial infarction, heart failure, and stroke. The risk varies depending upon the baseline cardiovascular event risk of the patient, the NSAID used, and its dose. Naproxen is generally preferred over other nonselective NSAIDs as it appears to be an exception in terms of significant cardiovascular toxicity. Of note, the beneficial effect of aspirin may be attenuated by prior or ongoing administration of some nonselective NSAIDs, including ibuprofen and naproxen. It is currently uncertain how the cardiovascular toxicity of NSAIDs will affect their chemoprevention application, if any. Interestingly, cardiac toxicity during CRC chemoprevention trials was the main reason for withdrawing COX-2-specific inhibitors [67].

Conclusions

The data reviewed above strongly indicate the following: the chemoprevention of CRC by NSAIDs is feasible; conventional NSAIDs alone are not suitable for CRC chemoprevention due to low efficacy and high toxicity; certain subpopulations at risk for CRC may be better candidates for NSAID-based chemoprevention; targeting early stages of carcinogenesis such as ACF holds promise; and some of the newer modified NSAIDs may turn out to be far superior to their parent compounds, thus being appropriate for chemoprevention.

These conclusions raise a multitude of issues. As perhaps in all of medicine, the common element in trying to deal with them is that one must weigh the risk and benefit of each contemplated intervention. Chemoprevention requires administration of an agent for prolonged periods of time. The higher the risk of CRC in a given individual and the lower the cumulative side effects of the chemopreventive agent, the clearer is the justification for prescribing it. Conventional NSAIDs have side effects that increase with prolonged administration, especially in older patients who, because of comorbidities, take medications that interact with the chemopreventive agent. Thus, to justify the lifetime administration of a conventional NSAID, the CRC risk must be much higher than the <5 % probability that an individual at average risk has to develop CRC [68].

These and similar considerations suggest the need to focus on the transition of chemoprevention from its current investigational stage to that of clinical application. Relevant to it would be the following:

  • Identification of subjects at high risk and those most likely to benefit from the chemopreventive agent. The development of approaches to reliably estimate risk and of predictive biomarkers would be extremely helpful in this context.

  • Optimization of the administration of the chemopreventive agent. The work of Takayama and Niitsu has opened an important window into this aspect, as it suggests that attempting to arrest colon carcinogenesis at a very early stage may require relatively short periods of agent administration. Thus, the timing, dose, and duration of NSAID administration could be fine-tuned so that the lowest possible NSAID dose is used for the shortest period of time. Furthermore, in those at risk, it may be helpful such an intervention to commence at a relatively young age.

  • New agents or combinations of agents. The promise of higher efficacy and safety of the modified NSAIDs, such as PS, indicates that their further exploration may have merit. The efficacy of sulindac/DFMO supports the argument for combination approaches that will increase efficacy beyond the currently low levels of single agents.

  • Agents that may prevent other diseases in addition to CRC would shift dramatically the risk/benefit ratio in favor of the intervention. For example, preventing both colon cancer and heart attacks, as conventional aspirin does albeit weakly, is altogether different than only preventing one of these two diseases.

The enormous accomplishments of the last two decades have provided proof-of-concept for NSAID prevention of CRC and moved the field to its current exciting stage where rationally targeted implementation strategies can be contemplated. Although significant work lies ahead, there is ample reason for serious optimism.