Chemoprevention in Gastrointestinal Cancers: Current Status


  • Maria V. Grau,

    1. Department of Community and Family Medicine, Dartmouth Medical School and
    Search for more papers by this author
  • Judy R. Rees,

    1. Department of Community and Family Medicine, Dartmouth Medical School and
    2. Norris Cotton Cancer Center, Dartmouth-Hitchcock Medical Center, Lebanon, NH, U.S.A.
    Search for more papers by this author
  • John A. Baron

    Corresponding author
    1. Department of Community and Family Medicine, Dartmouth Medical School and
    2. Norris Cotton Cancer Center, Dartmouth-Hitchcock Medical Center, Lebanon, NH, U.S.A.
    Search for more papers by this author

Author for correspondence: John A. Baron, Biostatistics and Epidemiology, Evergreen Center, 46, Centerra Parkway, Lebanon, NH 03766, USA (fax +1 603 650 3473, e-mail


Abstract: Chemoprevention, pharmacological intervention for disease prevention, aims to intervene in pathways that lead to clinical disease before the disease occurs. Cancer chemoprevention is a relatively new field, but for gastrointestinal cancers, clinical trials have highlighted the chemopreventive potential of several agents. For colorectal neoplasia, trials with aspirin and other non-steroidal anti-inflammatory drugs (NSAIDS) and calcium have demonstrated the most significant reductions of risk. In observational studies, NSAIDs also consistently appear to protect against oesophageal and gastric cancer. Calcium, and perhaps vitamin D, are also promising and have the advantage of being inexpensive, safe interventions. For the prevention of oesophageal cancer, antitumour-B and retinamide have provided hopeful results, although it is not clear that these findings can be extrapolated from the study populations in Asia to western countries. Evidence from China suggests that a combination of β-carotene, α-tocopherol and selenium may protect against oesophageal cancer, but the relative importance of each agent is unclear, and, again, their effects in other populations has not yet been assessed. Mass immunization against hepatitis B seems to be the most effective means of reducing the incidence of hepatocellular cancer worldwide. In addition, treatment with interferon α in patients chronically infected with hepatitis C virus shows considerable promise, given the increasing prevalence of hepatitis C virus carriage in recent years. TJ-9, polyprenoic acid and anti-aflatoxin compounds are also possible avenues that deserve future research.

Primary prevention aims to prevent disease before it occurs, e.g. the prevention of hepatocellular carcinoma by vaccination against hepatitis B virus. Secondary prevention aims to prevent the recurrence of a disease that has already occurred, as would be the case for prevention measures taken in patients with colorectal adenomas, precursors of a frank cancer. Pharmacologic intervention for disease prevention, chemoprevention, aims to intervene in pathways that lead to clinical disease before the disease occurs. This chemoprevention is a form of primary prevention. However, in practice, chemoprevention of invasive cancers may involve primary or secondary prevention of precursor lesions.

Although cancer chemoprevention is relatively new, there is a long history of chemoprevention in cardiovascular disease, for which interventions to counter hypertension, hypercholesterolaemia, and thrombosis have been used for decades to prevent clinical atherosclerotic disease. As in cardiovascular disease, cancer chemopreventive agents may be identified through epidemiological observations, laboratory experiments, and clinical experience with treatments for cancer itself. All these modalities have identified successful cancer chemoprevention agents – as well as some false leads. The decision to test putative chemopreventive agents depends on the accumulation of evidence from these sources, weighed against the possible side effects of the agent itself.

Epidemiological studies have identified many of the candidate agents that have been tested in gastrointestinal cancer chemoprevention trials. Randomized, double-blinded trials are ideally the best way to test the hypothesized benefits of a nutritional agent or a commonly used medication. These trials have the advantages of statistical equalization of the distribution of confounders between intervention arms by randomization; accurate determination of end-points by prospective follow-up; and the minimization of bias by concealing the nature of the intervention by blinding. Their disadvantages include expense and labour-intensive follow-up. Observational studies are limited by the biases that can result from uncontrolled confounding, as well as difficulties measuring exposures and outcomes.

In this review, we provide an overview of the major proposed chemopreventive agents against gastrointestinal tumours, focusing on those that have been evaluated in randomized clinical trials. Since our aim is to explore evidence of the usefulness of chemoprevention against gastrointestinal cancer, we have considered both primary and secondary prevention trials.

Colorectal cancer

Colorectal cancer is the best studied of the gastrointestinal cancers with regard to chemoprevention. For several reasons, research on the prevention of colorectal carcinogenesis has often focused on colorectal adenomas. This end-point avoids the size and complexity required for trials of colorectal cancer itself. Patients with sporadic adenomas are routinely followed with endoscopy, and patients with familial adenomatous polyposis receive even more intensive surveillance. Assuming that the effects of chemopreventive agents on adenomas reflect those on cancer, these end-points provide a convenient surrogate for study of the prevention of colorectal cancer itself (Peipins & Sandler 1994).

Antioxidants Antioxidants were the first agents evaluated in the chemoprevention of colorectal cancer, following the observation that colorectal cancer seemed to be less common in populations whose diets were rich in fruits and vegetables (Steinmetz & Potter 1991). The antioxidant vitamins, especially β-carotene, vitamins E and C and α-tocopherol, appeared to be perfect candidates to prevent colorectal cancer, given their minimal toxicity and their repeated association with a lower risk of colorectal cancer in several studies examining dietary intake or serum levels (Longnecker et al. 1992; Bostick et al. 1993; van Poppel 1993). Experimental studies have shown that antioxidants may have effects on cell proliferation and DNA mutation, mechanisms that are directly implicated in the formation of colorectal neoplasia (Paganelli et al. 1992).

However, despite this promising prelude, clinical trials failed to confirm a chemopreventive role for the antioxidants. The association of β-carotene with colorectal adenomas has been evaluated in three adenoma trials without any evidence of a benefit. Vitamin E also seems to be ineffective, although one large trial with a cancer end-point implied a protective effect. For vitamin C, an early adenoma trial suggested some benefit that subsequent studies failed to confirm. The antioxidant trials are summarized in (Sporn & Lippman 2003).

However, the interpretation of adenoma trials is not entirely straightforward. They study relatively early phases of carcinogenesis (the end-point adenomas are small, typically less than 5 mm). If antioxidants interfered with later stages of carcinogenesis, a protective effect would be overlooked. Moreover, the adenoma trials have lasted no more than four years, which could conceivably not be long enough for a chemopreventive effect to emerge. Nonetheless, several trials of antioxidants and lung cancer have been completed and are large enough (and long enough) to accrue sufficient numbers of colorectal cancer cases for meaningful analysis. The lack of association between β-carotene and colorectal cancer in these trials was consistent with the results of the colorectal adenoma trials (Anonymous 1994; Clark et al. 1996; Hennekens et al. 1996; Omenn et al. 1996).

Selenium, although not itself an antioxidant, is often considered in this group because of its important role in glutathione peroxidase, an antioxidant enzyme system. In the Skin Cancer Prevention Trial (Clark et al. 1996), selenium unexpectedly showed protective effects against cancers of the colorectum, lung and prostate. However, it is not clear whether these were chance findings or a reflection of genuine efficacy.

Dietary fiber Research has also focused on fiber as a possible alternative to explain the inverse association between risk of colorectal neoplasia and intake of vegetables and fruits (World Cancer Research Fund & American Institute for Cancer Research 1997). Fiber increases the motility of the large bowel and may dilute carcinogenic metabolites, actions that potentially might inhibit chemically induced carcinogenesis (Reddy 1999). Experimental studies in animals have broadly supported the idea that some sorts of fiber might protect against colorectal neoplasia. Epidemiologic studies, however, have been conflicting (Lawlor & Ness 2003). The evidence from clinical trials for the chemopreventive properties of fiber is not strong. Trials that used low fat/high fiber dietary interventions (McKeown-Eyssen et al. 1994; Schatzkin et al. 2000) showed no difference in adenoma recurrence between treatment groups, and studies that relied on cereal fiber supplements came to more or less negative results. One study, conducted among patients with familial adenomatous polyposis, found suggestions of a benefit that were not statistically significant (DeCosse et al. 1989); two others, conducted among patients with sporadic adenomas, were essentially negative (MacLennan et al. 1995; Alberts et al. 2000). Another trial showed increased risks related to supplementation with psyllium (Faivre et al. 1999).

However, it is not easy to place these negative findings into context. Soluble and insoluble fibers can have different effects on the bowel, and the fact that different fiber moieties were used in the trials means that there is less replication than appears at first glance. As discussed above, the duration of the trials may also have been insufficient to completely rule out a protective effect, and an impact on later stages of carcinogenesis is also conceivable, though admittedly unlikely.

Calcium and vitamin D. The suggestion that calcium intake might lower the risk of colorectal neoplasia has generated considerable investigation (Shaukat et al. 2005). The proposed mechanism involves calcium's ability to bind and precipitate soluble fatty acids and bile acids in the bowel lumen, inhibiting their well known proliferative and carcinogenic effects (Lamprecht & Lipkin 2003). In support of this hypothesis, animal studies have shown a protective effect of dietary calcium on bile-induced mucosal damage or experimental bowel carcinogenesis. Recently, it has been postulated that calcium affects cellular proliferation and differentiation by acting directly on the calcium-sensing receptor, a cell surface receptor expressed in normal and cancer cell lines.

The results of human epidemiologic studies have not been as consistent, perhaps reflecting the difficulties of dietary epidemiology (Sandler 2005). The two clinical trials of calcium supplementation have both shown a risk reduction. In the Calcium Polyp Prevention Study, supplementation with 1200 mg daily for 4 years resulted in a statistically significant 19% reduction in adenoma risk (Baron et al. 1999). This modest benefit was substantially larger (35%) when the end-point was advanced adenomas (Wallace et al. 2004). Although the European trial reported similar results for all adenomas, the 25% reduction in risk associated with calcium supplementation was not statistically significant in this smaller study (Bonithon-Kopp et al. 2000).

Along with calcium, the anticarcinogenic properties of vitamin D have been extensively reported. The vitamin D receptor is a nuclear-ligand activated receptor that is expressed in almost all tissues in the body; this widespread distribution indicates a role for vitamin D beyond calcium homeostasis. In fact, experimental studies have repeatedly shown that vitamin D can inhibit cell proliferation, induce differentiation and promote apoptosis, processes directly involved in the development of human carcinogenesis. Animal and observational data have consistently confirmed that vitamin D protects against colorectal neoplasia (Lamprecht & Lipkin 2003). The best evidence to date of a protective effect of vitamin D was provided by a study that used data from the Calcium Polyp Prevention Study (Grau et al. 2003). This study showed that, among participants with serum levels of 25-OH vitamin D above the median (29.1 ng/ml), calcium supplementation conferred a statistically significant reduction in adenoma risk of almost 30%, but it had no such effect among those with serum vitamin D levels below the median. Within the calcium treatment group, there was a risk reduction of 12% for each standard deviation increase in serum vitamin D. These results provide evidence of an interaction between calcium and vitamin D, which suggests that they act together, rather than independently, in their protection against colorectal neoplasia.

Non-steroidal anti-inflammatory drugs (NSAIDs). Non-steroidal anti-inflammatory drugs were first shown to decrease colorectal neoplasia in animal studies (International Agency for Research on Cancer 1997). Many believe they do this by inhibiting cyclooxygenase enzymes, thereby preventing the synthesis of prostaglandins and other metabolites that are associated with inflammatory processes and carcinogenesis (Herendeen & Lindley 2003). Positive findings from these experimental studies motivated numerous human epidemiological studies. The great majority of these have pointed to a chemoprotective effect, suggesting that risk can be reduced by about 50% among individuals who take aspirin or other NSAID's regularly (International Agency for Research on Cancer 1997). Two clinical trials in patients with familial adenomatous polyposis have found that sulindac can lead to polyp regression and prevention of new polyps (Labayle et al. 1991; Giardiello et al. 1993). Three other trials using aspirin have found a protective effect on sporadic adenomas. The Aspirin Polyp Prevention Study (Baron et al. 2003) reported a reduction of adenoma risk of 17% with aspirin 81 mg, and an even larger risk reduction for advanced adenomas (47%). However, the adult dose, aspirin 325 mg did not appear to have any effect on adenoma risk in this study. In contrast, the Colorectal Adenoma Prevention Study (Sandler et al. 2003) and the French Study (Benamouzig et al. 2003) found reductions of 35% and 27% respectively for all adenomas with a similar dose, but there was no effect on advanced adenomas. Two clinical trials of aspirin have reported colorectal cancer end-points, but neither of these continued aspirin treatment long enough for an effect on cancer (as opposed to the earlier adenomas) (Giovannucci et al. 1994; Cook et al. 2005). Thus, anti-inflammatory agents show considerable promise for the chemoprevention of colorectal neoplasia. The mechanisms underlying this effect are not clear; alternative mechanisms to cyclooxygenase inhibition have been postulated recently and these need to be explored in further detail (International Agency for Research on Cancer 1997; Smith et al. 2000).

Ursodeoxycholic acid. Several experimental studies have shown that ursodeoxycholic acid can prevent colon carcinogenesis (Alberts et al. 2005). The chemopreventive actions of this drug are related to its ability to inhibit the mitogenic signaling from secondary bile acids and thus to suppress cell growth. Observational studies have shown an inverse relation between ursodeoxychlolic acid treatment and colorectal neoplasia in patients with ulcerative colitis and primary sclerosing cholangitis (Tung et al. 2001; Pardi et al. 2003), and among patients with primary biliary cirrhosis (Serfaty et al. 2003). One small clinical trial corroborated these findings: the risk of colorectal neoplasia decreased by more than 70% over a median treatment with ursodeoxychlolic acid of 6 years in patients with ulcerative colitis, a risk factor for colorectal cancer. Another larger trial found a significant 39% risk reduction for recurrence of highly dysplastic adenomas over a three year period, although the effect on total adenoma recurrence was not statistically significant (Alberts et al. 2005). A third randomized trial among patients with previously removed colorectal adenomas or early-staged adenocarcinomas showed no difference between treatment with placebo and ursodeoxychlolic acid after one year of follow up (Larson et al. 2002). More research would be valuable to clarify these suggestive findings.

Hormone replacement therapy. The association of hormone replacement therapy in women and the risk of colorectal cancer have been explored in two large meta-analyses that summarized findings from observational studies conducted in several countries studies and involving a variety of hormone replacement therapy products (Hebert-Croteau 1998; Grodstein et al. 1999). The beneficial effects reported here were corroborated later in the Women's Health Initiative trial, where combined oestrogen and progestin or placebo were randomly assigned to healthy women (Rossouw et al. 2002). In this trial, a significant 37% reduction in risk of colorectal cancer was found in the treatment group. However, the use of oestrogen alone in another trial from the Women's Health Initiative did not show any significant protective effect (Anderson et al. 2004). To date, the mechanisms behind these findings are not clear, and the well-known risks of hormone replacement therapy regimens in relation to other type of cancers (Warren 2004) have to be taken into consideration before other trials can address these associations.

Oesophageal cancer

Oesophageal cancer has not been as well studied as colorectal cancer, in part because it does not have a surrogate end-point comparable to colorectal adenomas, and also because the incidence in developed Western countries is relatively low. Much epidemiologic research on this cancer has been centered in Linxian, China, where the incidence and mortality rates are the highest in the world.

Cyclooxygenase-2 expression is elevated in dysplastic oesophageal mucosa (Yu et al. 2003). Furthermore, NSAID use has been associated repeatedly with a reduction in risk of oesophageal cancer in observational studies. A meta-analysis found a significant inverse association with use of NSAIDs, that was stronger for squamous carcinoma than for adenocarcinoma and also stronger for aspirin than other NSAIDs (Corley et al. 2003).

Two randomized nutritional intervention trials in Linxian, China, have also explored the associations between oesophageal cancer and the use of different mixtures of vitamins and minerals, including β-carotene, α-tocopherol, ascorbic acid and selenium. One trial included only high risk subjects with a cytological diagnosis of oesophageal dysplasia (Li et al. 1993); the other included people from the general population (Blot et al. 1993). None of the combinations of agents significantly reduced the incidence of cancers. However, pre-trial serum levels of selenium in the highest quartile were associated with a 44% lower risk of oesophageal cancer (Mark et al. 2000), a finding that merits further investigation.

Also in China (Lin et al. 1990), a study of high risk patients with oesophageal dysplasia reported a reduction in oesophageal cancer of approximately 50% among those treated with antitumour-B and 28% among those treated with retinamide.

Gastric cancer

As for oesophageal cancer, data concerning the prevention of gastric cancer are more limited than those for colorectal cancer. Infection with Helicobacter pylori is a recognized risk factor for gastric cancer (International Agency for Research on Cancer 1994), suggesting that eradication of infection might reduce cancer incidence. In observational studies, NSAIDs appear to exert preventive effects against gastric cancer, but clinical trials have not explored this association (Wang et al. 2003). Two nutritional trials carried out in Linxian, China, focused on cancer of the gastric cardia, which has a high incidence in this region. These trials showed a 21% reduction in the risk of cancer among subjects randomized to β-carotene-α-tocopherol and selenium (Blot et al. 1993). Another trial was conducted in Narinyo, Colombia, where gastric carcinoma and precursor lesions are also frequently diagnosed (Correa et al. 2000). This trial studied the effects of various interventions on pre-cancerous lesions present at baseline. The treatment arms included triple anti-helicobacter pylori therapy, ascorbic acid supplements, β-carotene supplements, and all possible combinations of these three interventions. In this study, all interventions resulted in significant regression of existing pre-malignant lesions in a confusing pattern that did not clearly distinguish the relative effectiveness of the individual agents.

Hepatocellular carcinoma

Liver cancer is one of the most common cancers worldwide, but has a wide geographic variation: Asia, the Pacific Islands and Sub-Saharan African countries have much higher incidence rates than Western countries. Hepatocellular cancer is closely associated with infection by hepatitis B and hepatitis C viruses. It has been estimated that hepatitis B virus infection accounts for 80% of all hepatocellular cancers. In developing countries where these viruses are highly prevalent, perinatal transmission and chronic infection from early childhood account for most cases of liver cancer (Moriwaki 2002).

Several studies have demonstrated the great value of immunization for the primary prevention of this cancer. In 1984, Taiwan started a nationwide hepatitis B vaccination program, initially only in newborns and later also in adults. The annual rates of hepatocellular cancer decreased dramatically, virtually eradicating liver cancer in children (Chang et al. 2004). A randomized study in Gambia, Africa, has shown that administration of the vaccine during the first year of life confers a high degree of protection against chronic carriage of hepatitis B virus, although it is too early to assess the ultimate effect on liver cancer (Montesano 2002).

An additional risk factor for hepatocellular cancer in areas with high prevalence of hepatitis B virus infection is the dietary consumption of aflatoxins. These toxins are produced by two fungi, Aspergillus flavus and Aspergillus parasiticus, which, under appropriate conditions of humidity and moisture grow and produce toxins in virtually any foodstuff. Synergism between hepatitis B virus and these toxins in causing hepatocellular cancer has been well documented (Kensler et al. 2004). Chemopreventive agents could potentially neutralize the effects of these toxins. Two agents have been tested with satisfactory results: Chlorophyllin, which binds aflatoxins and impairs their absorption and Oltipraz, an inducer of detoxificant enzymes in the liver (Kensler et al. 2004).

Another important risk factor for hepatocellular carcinoma is chronic infection with hepatitis C virus, which has been responsible for increasing rates of liver cancer in recent years and is now its major cause in developed countries (Heathcote 2004). Multiple retrospective studies and uncontrolled studies report an inverse association between interferon therapy and hepatocellular cancer in patients with hepatitis C virus with or without cirrhosis. To date, only one of several randomized trials has corroborated this finding (Chou et al. 2004).

Other possible strategies for secondary prevention in patients with established chronic cirrhosis and premalignant lesions involved herbal medicines and an acylic retinoid. In Japan, a trial was conducted among non-B cirrhotic patients of a Chinese herbal medicine, TJ-9, proven to have antitumour activity in laboratory studies; the trial showed a risk reduction of more than 40% (Oka et al. 1995). Also in Japan, a trial among patients with a history of hepatocellular cancer reported a reduced risk of relapse or new primary cancers when treated with polyprenoic acid (Muto et al. 1996), opening another line for future investigation.

Pancreatic cancer

To date, no chemoprevention trials have been conducted in pancreatic cancer. However, there are numerous experimental and epidemiologic studies that point to various drugs as potential protective agents. Dietary isoprenoids, somatostatin analogues, selective oestrogen modulators and anti-androgen agents have shown some chemopreventive potential in animal and in vitro studies (Wolff 2003). Aspirin, NSAIDs, and selective cyclooxygenase-2 inhibitors have also been proposed (Wolff 2003) although there are also some indications that aspirin may potentially have an adverse effect (Baron 2004). It will be very difficult to resolve this issue with clinical trials, given the relatively rarity of this end-point.


To date, the findings from clinical trials on the different gastrointestinal cancers have highlighted the chemopreventive potential for a variety of different agents and are the basis for optimism regarding the potential for effective clinical prevention in the future. For colorectal neoplasia, trials with aspirin and other NSAIDs and calcium have demonstrated the most significant reductions of risk. In observational studies, NSAIDs also seem to have a consistent protective effect against oesophageal and gastric cancer. Arguably, for the luminal gastrointestinal tract, NSAIDs offer the best hope for chemoprevention. Calcium (and perhaps vitamin D) are also very promising, and have the advantage of being inexpensive, safe interventions.

Clinical trials in Asia have identified other avenues for gastrointestinal chemoprevention. Antitumour-B and retinamide provided hopeful results for oesophageal cancer, although it is not clear that these findings can be extrapolated to other populations. The Linxian trial in China found that a combination of β-carotene, α-tocopherol and selenium conferred a significant reduction of risk. Whether this finding was somehow related to the particular combination, or to the particular population studied (with marginal nutritional deficiencies) is not clear. Although mass immunization against hepatitis B seems to be the most effective means of reducing the incidence of hepatocellular cancer worldwide, strategies to tackle other well known risk factors have been tested in clinical trials. Of all these, treatment with interferon-α in patients with chronic hepatitis C virus infection seems to be the most effective and promising, given the increasing prevalence of hepatitis C virus carriage in recent years. TJ-9, polyprenoic acid and anti-aflatoxin compounds are also possible avenues that deserve future research.