Article Text

Download PDFPDF

Chemoprevention of colorectal cancer
  1. P BOYLE
  1. Department of Medicine, Queen Elizabeth Hospital, Birmingham B15 2TH, UK
  2. Division of Epidemiology and Biostatistics, European Institute of Oncology, via Ripamonti 435, 20141 Milan, Italy

Statistics from

Colorectal cancer is the fourth commonest form of cancer in men with 678 000 estimated new cases per year worldwide, representing 8.9% of all new cancers. The disease is most frequent in Occidental countries and particularly so in North America, Australia, New Zealand, and parts of Europe. Prospects for colorectal cancer control are bright and a number of possible approaches could prove fruitful. Among these, pharmaceutical measures seem to be valid and logical approaches to the prevention of colorectal cancer and diminishing its impact. Such approaches could concentrate in primary prevention in at-risk subjects or be applied in altering the course of precursor or established disease. Treatments used must fulfil basic requirements of biological plausibility and safety in continued use in large numbers of subjects. Those available include vitamins and minerals, and other drugs with potential as antioxidants, immune modulators or promoters of cell differentiation or apoptosis. Of the various regimens suggested, vitamin A supplementation may even predispose to adverse outcomes, and antioxidant vitamins in general have no coherent body of evidence to support their use. N-acetylcysteine and ursodeoxycholic acid have promising characteristics but there are as yet no clinical data to support the use of the former in gut epithelial cancer, and formal dose ranging studies must be carried out before the latter is submitted to large scale trial. Folate shows promising characteristics but non-steroidal anti-inflammatory drugs and vitamin D seem the most promising agents. Both seem to reduce the incidence of disease, and to reduce growth rates and/or induce differentiation or apoptosis in gut epithelial cancer cells. Both are also well understood pharmacologically. They may be preferred to newer selective compounds in the same class until these newer compounds are confirmed as safe for widespread long term use.


Colorectal cancer is the fourth commonest form of cancer worldwide with an estimated 678 000 new cases diagnosed in 1985.1High incidence rates are found in Western Europe, North America and Australasia, intermediate rates in Eastern Europe with the lowest rates found in sub-Saharal Africa.2 In the United Kingdom there are some 25 000 to 30 000 new cases each year. The disease is not uniformly fatal, although there are large differences in survival according to stage of disease. It is estimated that there are, however, still nearly 400 000 deaths from colorectal cancer worldwide annually.3

Current strategies for colorectal cancer control have concentrated upon early detection, either as screening tests for faecal occult blood (such as haemoccult testing), or more recently by endoscopy, and are essentially directed at preventing progression of the polyp–cancer sequence, or at the detection of cancer before invasion and metastasis have occurred. Alternative strategies are required, aimed at primary or secondary prevention, and which are widely applicable. One route would be the refinement of molecular screening methods designed to focus attention upon individuals who seem to be inherently at high risk by reason of their genetic make up. However, we are at present ill-placed to apply such methods to the common sporadic variety of colorectal cancer. Moreover, the method would still concentrate upon early detection in susceptible people rather than upon general prevention of disease, or alteration of its course once ordinarily diagnosed.

In these circumstances it makes sense to consider whether direct preventive measures could be applied and, given the large costs, complexities and duration of study, what would be likely to be the most fruitful methods. This article reviews evidence on the value of a range of potential chemotherapeutic agents.

Methods to be considered include general population intervention, polyp and cancer prevention in at-risk subjects, and trials in patients treated for cancer, an area which could be described as adjuvant chemoprevention where a treatment is given not primarily to treat the tumour but to prevent or slow down the development of local recurrences, metastases or second primary tumours.

Choice of agent

Any agents chosen for study should be capable of long term administration and should be largely if not completely free of adverse effects. In addition, there should be plausible epidemiological and/or biological evidence in favour of ability to inhibit polyp development and/or retard tumour growth. Finally, it should be possible to combine their use with that of antimitotic drugs so that a single trial could consider adjuvant therapy as well as adjuvant chemoprevention. The requirement of adverse effect free treatment which can be given continuously for very prolonged periods makes novel chemical entities unpromising candidates for immediate widespread use.

General preventative measures which could be applied essentially divide into lifestyle alterations and direct chemopreventative treatments. Lifestyle alterations directed at, for instance, reducing saturated fat intake have obvious logic and appeal given the demonstrable relation between fat intake and large bowel cancer incidence, and the general public health benefits. However, evidence from studies of cardiovascular disease illustrates the difficulty of devising dietary advice schedules to which people are likely to adhere, and which could be applied in large simple trials.

There is obvious logic therefore in considering chemopreventative measures provided attractive candidate treatments exist. This review concentrates upon pharmaceutical measures because their simplicity makes them easier to apply than broad dietary measures such as increasing fruit and vegetable intake. In this context, measures designed to inhibit vascularisation of tumours are not considered in this review.

Candidates include, at least, vitamin A or β carotene, vitamin C, vitamin D, vitamin E, calcium supplements, folate and anti-inflammatory drugs and H2 antagonists. Relatively novel chemical entities such as protease inhibitors are at an earlier stage making prolonged treatment too speculative a possibility. Taken overall there are suggestions of benefit for all of the above compounds, some stronger than others, in reducing the risk of developing colorectal cancer, and/or in preventing polyp occurrence. Effects on secondary spread are little understood except that anti-inflammatory agents have been used clinically in the past in trying to inhibit osteolysis.

There are also extensive animal or observational studies on a range of other compounds which includes the ornithine decarboxylase inhibitor difluoromethyl ornithine, flavonoids, dithiolethionine inducing glutathione transferase activity, selenium, and others.

Use of anti-inflammatory agents

There is abundant evidence that the use of non-steroidal anti-inflammatory drugs (NSAIDs) is associated with reduced risks of colorectal cancer and adenomatous polyps.4-8 Effects have been demonstrated consistently, though not completely uniformly, are evident in case-control and cohort studies, and seem to be dose and duration of treatment related (table 1). Effects are biologically plausible because NSAID use seems to prevent or reduce the frequency of carcinogen induced animal colonic tumours,5 ,6 because NSAIDs seem to reduce growth rates in colon cancer cell lines, and because polyp formation in familial adenomatous polyposis coli (APC) seems to be retarded.11 ,12

Table 1

Risks of colorectal cancer in relation to aspirin or NSAID use

NSAIDs could act by switching off mitogenic stimuli, by inducing differentiation, or apoptosis or through enhancing immune surveillance. Clear evidence to support a particular mechanism as being of dominant importance is lacking. However, the most obvious mechanism is through inhibition of cyclo-oxygenase (COX). COX activity, particularly COX-2, is upregulated in colorectal cancers but is less easily detectable in colorectal adenomas or in normal colonic epithelium. Further strong evidence for the significance of COX-2 expression comes from studies of the MIN mouse, which carries the APC gene, and develops multiple intestinal adenomata spontaneously. If animals with the mutation are crossbred with others with a disrupted COX-2 gene then it can be shown that effects on the induction of COX-2 activity are associated with altered rates of adenoma formation. Administration of the NSAID piroxicam to MIN mice reduces adenoma development, as might be expected. The relevance of APC gene studies to the common variety of colon cancer is emphasised by evidence that APC mutations are commonly detected in colonic carcinomata.13-18

There is also good evidence to suggest that induction of apoptosis is important, effects being demonstrable in cell lines and in familial adenomatous polyps.19-21 A possible link with COX inhibition derives from data indicating that COX-2 activity affects the occurrence of apoptosis in rat intestinal epithelial cells.22 The apparent simplicity of these inter-relations is, however, disturbed by results indicating that sulindac sulphone, which is not a significant inhibitor of COX, as well as the sulphide, a conventional NSAID, are both effective suppressants of growth in tumour cell lines.23

Furthermore, the aspirin metabolite salicylate, which is not a prominent COX inhibitor, seems to induce apoptosis in transformed colonic cell lines. Sensitivity seems to be increased in the later stages of neoplastic progression.23

Taken overall, the evidence indicates that NSAIDs are effective suppressants of tumour cell proliferation, can prevent or reduce the occurrence of experimentally induced colonic epithelial tumours, and that in humans use of NSAIDs for other therapeutic purposes reduces the frequency of colorectal cancer. Cases can therefore be made for considering NSAIDs in preventing cancer development, or in altering the behaviour of diagnosed disease as adjuvant therapy in preventing or retarding recurrent disease. Use of conventional NSAIDs is attractive because they have been licensed for human treatment for many years and their effects, outside cancer, are well understood. Newly developed COX-2 inhibitors are inherently attractive because their selectivity could prevent the well known upper gastrointestinal adverse effects of conventional NSAIDs. However, if COX inhibitors were to be used in preventing or altering the behaviour of colorectal cancer then it is likely that treatment would have to be given for many years. Furthermore, the expectation might be of modest alterations in disease behaviour so that benefits of treatment would be limited to a small proportion of individuals. In these circumstances it is doubtful whether a newly developed COX inhibitor could be used until its clinical pharmacological profile and pattern of adverse responses were well characterised.

Outside primary prevention of colorectal cancer, where trials in subjects at high risk seem attractive, there is a good case for treating those with diagnosed disease who are at risk of poor outcomes through worsening of existing disease or recurrence.

Choice of drug

Use of sulindac is supported by its effects in retarding adenoma development in familial adenomatous polyposis12 (though maybe not in sporadic disease24), and by data showing that it can induce apoptosis—although this is likely to be a class effect.

Indomethacin or other COX inhibitors have been suggested to improve natural killer cell activity, and to enhance non-specific immunotherapy of experimental lung tumour metastases.25-27 In a randomised study of cervical cancer in 160 patients, use of indomethacin improved survival of those given radiation treatment by 27% at five years and 41% at 10 years.28 Other studies in colorectal cancer do not seem to have been undertaken. Salicylate as sodium salicylate would be an interesting alternative, but it is unclear if it would have any therapeutic advantages, and dosage could be limited and compliance reduced by symptoms of salicism, vertigo, tinnitus, and deafness. Conventional NSAIDs would therefore probably be preferable to simple salicylate.

NSAIDs have class adverse effects on the kidney (interstitial nephritis), skin (rash and photosensitivity), lung (predispose to asthma), and liver (hepatitis—particularly with diclofenac). However, none of these, individually or collectively, is as frequent as gastrointestinal bleeding from peptic ulcers and, to a lesser extent, from the colon. Risks vary up to 20-fold between agents and by up to 10-fold by dose (table 2).29

Table 2

Risks of peptic ulcer bleeding in subjects taking commonly used non-aspirin NSAIDs for at least three months,29 and in individuals taking prophylactic aspirin at least five times a week for the past39

There is no obvious way in which risks of gastrointestinal adverse effects can be prevented, except by reducing doses or by giving concurrent antisecretory or gastroprotective treatments. The use of misoprostol as a gastroduedenal protective drug30 would seem to be unwise because deliberate addition of prostaglandin could perhaps interfere with the antagonist and possibly the desired therapeutic effects of an NSAID. The addition of histamine H2 antagonists might be expected to reduce, depending on dose and choice, the frequency of gastric and duodenal ulcer31 ,32 and possibly the rate of ulcer complications, though this is unproved. Use of a proton pump inhibitor might be more beneficial through more powerful antisecretory effects but with the theoretical possibility of adverse tumour responses through raised serum gastrin concentrations, although these have been disputed.33-35 A more important practical disadvantage of antisecretory therapy would be through addition of an extra treatment, so reducing the likelihood of overall compliance.

If epidemiological findings of possible preventative actions are a fair guide then available evidence (table 1) suggests that aspirin may, if anything, be more effective in preventing disease and so perhaps, if the further extrapolation is fair, in modulating existing disease behaviour. However, it should be noted that experimental and clinical evidence largely relates to non-aspirin NSAIDs.

Epidemiological evidence gives limited guidance on what doses of aspirin might be used deliberately to prevent colorectal cancer occurrence because retrospective enquiry is likely to be a poor method of determining what doses patients habitually took many years before. Prospectively obtained data in the Nurses Health Study indicated that 10–15 years is required to see an effect.36 The US physicians study, which suggested that 325 mg aspirin on alternate days might be relatively ineffective is not incompatible in findings because its time span (seven years) was shorter.37 Doses of 325 mg daily, which would raise the risk of ulcer bleeding by about fourfold (table 2), would seem necessary therefore in prevention and maybe also, if extrapolation is justified, in adjuvant treatment of established disease. Aspirin has additional potential cardiovascular benefits through inhibiting platelet COX, although similar benefits might be obtainable with—for example, regular indomethacin use. Note here that colon cancer incidence rates tend to be particularly raised in countries with high mortality rates from cardiovascular disease so that the potential for added benefit may be greater than expected.

Enteric coated aspirin has two possible advantages. It might deliver aspirin direct to the colon and it might reduce the risk of upper gastrointestinal adverse effects. Both advantages seem doubtful. Firstly, enteric coated drug is likely to be largely if not completely absorbed in the small bowel and, secondly, current evidence suggests that enteric coating may not greatly, or at all reduce the risk of ulcer complications.38 ,39

The selective COX inhibitors may ultimately prove to have a place, but judgment must be deferred until their pharmacological profiles and long term effects, good and not so good, are better understood.

Vitamin A or betacarotene and retinoids

Retinoids and retinoid related molecules show promise as chemotherapeutic or preventative agents because they can induce differentiation and apoptosis, the differentiating effects being well demonstrated in myeloid cells.40 ,41

Interpretation of evidence is complicated because the generic term vitamin A can be used to refer to the preretinoid betacarotene and to retinol. Furthermore, the retinoids include compounds, such as all trans and 9 cis retinoic acid, which are clearly mutagenic, which limits therapeutic value. Complications are reinforced by evidence that retinoids can have notably dissimilar effects experimentally according to the system in use. Thus 9 cis retinoic acid can either enhance or inhibit the effects of 1,25 dihydroxy vitamin D in retarding colon cancer cell line growth.42 Differences between retinoids will depend greatly on variable activation of retinoic acid (RAR) and retinoid X (RXR) receptors.43

Low serum betacarotene concentrations have been noted to be associated with later liability to all cancer, particularly lung cancer and also stomach, but not colon, cancer.44 Antioxidant vitamins (betacarotene and C, or A, C and E) were however found elsewhere to be protective against rectal cell proliferation in patients with rectal adenomas45 ,46 and in the prevention of recurrence of adenomas.47

By contrast, although a contemporaneous study of micronutrient intake suggested protection associated with intake of betacarotene, ascorbic acid, vitamin E, and folate,48 a review of five dietary studies showed unconvincing evidence of protection against colon cancer by vitamin A.49 Furthermore, a large (three times as large as the positive report47) study of prevention of adenoma recurrence using betacarotene 25 mg daily, instead of vitamin A 30 000 IU daily, together with vitamins C and E showed no evidence of protection (see figures below under vitamin C).50 Pure retinoids may ultimately prove to be of interest if consistent inhibitory effects on human colon cancers can be shown, and if their safe use is demonstrable.

Three large, randomised intervention studies of betacarotene have now produced results which were consistent with either no effect or a deleterious effect on lung cancer risk.51-53 Taken overall, the evidence to support use of vitamin A or betacarotene in the prevention of colorectal cancer or modulation of its behaviour is weak with indications, if anything, of possible harm in other pathology.

Other retinoids being explored include 13 cis retinoic acid and 4 hydroxyphenretinamide.54 ,55 In general all compounds present potential problems because of adverse responses. Those for isotretinoin, an isomer of tretinoin, include corneal opacities, cataracts, convulsions, myalgia, thrombocytopenia, and neutropenia as well as fetal malformations. There is, however, reasonable prospect of obtaining retinoid molecules which increase differentiation and/or induce apoptosis because synthetic molecules can be constructed which activate limited parts of the complex retinoid response network. At least in myeloid cells retinoic acid and anti-inflammatory agents can act on a single intracellular target site.56

Vitamin C

Generally dietary studies have suggested that high fat intake is associated with increased risk of colon cancer and that consumption of fruit, vegetables and fibre might be protective. Reduced risk of colon cancer has been associated with the use of vitamin C,48 and vitamin C combined with vitamin A seemed to reduce crypt cell proliferation rates in humans measured by the bromodeoxyuridine technique. Similar effects were noted with a diet supplemented with vitamins A, C and E when using the tritiated thymidine technique.45 ,46 Finally, protection against polyp formation using vitamin C alone was noted in one study of familial polyposis.57

The assumption has been of an anti-oxidant mechanism. Vitamin C as part of such a regimen (vitamin A 30 000 IU daily, vitamin C 1 g daily and vitamin E 70 mg daily) appeared to be protective in a study of 255 patients with polyps randomised to three treatment arms (vitamins, lactulose or placebo). Recurrence rates in the 209 evaluable subjects were 5.7, 14.7 and 35.9% respectively.47 By contrast, another much larger study included 864 patients with 751 completing one to four years’ treatment. In the 751 completing treatment 36% developed polyps. There was no advantage to combined vitamin treatment (relative risk 1.08, 95% confidence interval (CI) 0.91 to 1.29).50 Two other studies, of vitamins C and E combined, one of recurrence of familial polyposis, and the other of 143 patients for recurrence of simple polyps gave negative results.58 ,59

The evidence favouring vitamin C and/or anti-oxidant therapy in general is not strong, the results from the single large polyp study being particularly unfavourable.

Vitamin D

This vitamin in its activated dihydroxy form plays a central role in cell differentiation. People with low intake of vitamin D or with low serum concentrations have been noted by some to be at increased risk of colon cancer,60 ,61 and in North America, albeit from an inherently weak study design, an inverse correlation between sunlight exposure and liability to colon cancer has been described.62 Animal studies also suggest protective effects against experimental (e.g. 1–2 dimethyl hydrazine induced) colon carcinogenesis.63 ,64 1,25 dihydroxy vitamin D3 (OH D3) at concentrations of 10−10 molar inhibits multiplication within colon cancer cell lines and induces maturation. Functional vitamin D receptors are consistently demonstrable in human colorectal neoplasms,65 and human tumour xenograft growth is slowed by treatment with OH D3.66

The effects of OH D3 on cell differentiation seem likely to be mediated by a non-genomic cell membrane pathway.67 Thus protein kinase C (PKC), which plays a critical part in signal transduction within cells in the control of differentiation and growth, seems to be directly activated by OH D3. In this context it is noteworthy that at least one isoform, PKCe, which is calcium independent, is activated by OH D3 and OH D3 effects on cell differentiation can be mimicked by a non-hypercalcaemic analogue,68 ,69 and a non-calcaemic analogue inhibits azoxymethane induced colonic tumorigenesis, possibly at an initiation stage through effects on K-rasmutations.70

Evidence is not entirely one-sided. Thus two dietary studies showed weak evidence of protection, the one suggesting benefit from supplements but not from dietary vitamin D, and the other a non-significant odds ratio (0.73, 95% CI 0.45 to 1.18),71 ,72 compatible nevertheless with useful effects not clearly demonstrable in a study lacking power.

Taken overall, the evidence is reasonably in favour of vitamin D having useful antitumour activity, but there is a lack of clinical data in support. This is likely to be because vitamin D has well known adverse effects of hypercalcaemia which may have been a limitation to its past consideration. The use of modest doses, while ensuring that milk or calcium intake are not high, would however seem safe. Clues to the likely safe levels could be obtained from studies of vitamin D treatment in the elderly.73 As with COX inhibitors there is evidence, if somewhat less strong, of actions in preventing tumour occurrence and in altering the behaviour of established disease, but possible inter-relations do not seem to have been studied.

The attractions of the selective COX-2 inhibitors are mirrored by those of the non-hypercalcaemic analogues of vitamin D as promising preventative or adjuvant treatments once long term safety is confirmed.


Calcium was proposed on theoretical grounds as having a potential role in preventing colorectal carcinogenesis through reducing bile acid concentrations in the bowel. Effects in reducing cell proliferation have been detected in human studies and in animals, but have been denied by others.74 ,75 Whether any actions in fact arise through inactivating bile acids rather than through the vitamin D pathway is unclear. A better case can probably be made for vitamin D itself.

Vitamin E

Findings in five studies independently suggested non-significant trends towards reduced blood concentrations of α-tocopherol in subjects subsequently developing colorectal cancer compared with controls. Combination of the data showed a barely significant change in odds ratio 0.6 (95% CI 0.4 to 1.0) for the highest quartile of serum concentration compared with the lowest, but reduced to 0.7 (95% CI 0.4 to 1.1) when allowance was made for serum cholesterol concentrations.76 Other data (Iowa Women’s Health Study) suggested a highly significant inverse trend between serum concentration and cancer risk (p for trend <0.001).77 A third study suggested protection from vitamin E as well as folate, ascorbate or betacarotene.48

Use of vitamin E has the advantage that it is, for all practical purposes, safe. A mechanism of action is unclear beyond enhancement of anti-oxidant concentrations, the value of which is unclear. However, such a regimen failed to reduce adenoma incidence (when given with ascorbic acid and betacarotene50) although another smaller study—see vitamin C section—using 70 mg daily with vitamins C and A, suggested highly significant protection.47


Dietary case-control comparisons show protection from colon cancer associated with the consumption of fruit and vegetables, and after adjustment for other factors, folate emerges as potentially protective.78–80 Support comes from experimental tumour studies, with folate depletion increasing the yield and repletion diminishing the yield.81 81a DNA hypomethylation occurs in colonic tumours, and lack of folate reduces methyl group availability for DNA methylation which itself is associated with gene silencing.82 Studies indicating raised risks of colon cancer or adenoma in association with alcohol intake78 82aare consistent because demand for folate is raised by alcohol consumption. In ulcerative colitis the risk of dysplasia also seems to be reduced in those with higher levels of red cell folate.83 Studies of folate are reasonably consistent in indicating benefit. A basis in supplying methyl groups to convert homocysteine to methionine and then for DNA methylation via S-adenosylmethionine83a is logical. However, if this is true it is unclear why evidence in support of methionine intake is not stronger, whereas protein intake, as least as red meat, seems if anything to be a risk factor.


This compound has come under consideration because of anti-oxidant properties, and because its use appears to reduce the frequency of experimentally induced lung and skin tumours.84 There are no studies of human gastrointestinal cancer treatment, but a large European study (Euroscan) using N-acetylcysteine and/or retinol palmitate in a factorial design treating patients with resected oral, laryngeal or lung cancer is nearing completion.85

Ursodeoxycholic acid

Treatment with this, the 7-B-epimer of chenoeoxycholic acid, reduced the incidence of experimentally induced colon tumours in rats. Administration of ursodeoxycholic acid has also been shown notably to reduce the concentration of the secondary bile acid deoxycholic acid in the colon, while increasing that of lithocholic acid, although both have been considered to be promoters of colon cancer.86 ,87 Whether anti-cancer effects are by this route or another, notably modulation of protein kinase C isoforms,88 is unclear. Applicability in cancer prevention is uncertain, thus the dose required and the critical timing would have to be established.

H2 antagonists

Although it was initially feared that H2 antagonist treatment would precipitate the occurrence of gastric cancer through allowing bacterial overgrowth in the stomach and conversion of dietary nitrate to nitrosamines, supportive evidence for this view has not emerged. On the contrary there is evidence that use of H2antagonists can improve the course of gastric cancer, and some evidence that treatment is beneficial in colorectal cancer.

In a Danish study89 181 patients with gastric cancer were randomised to receive cimetidine 1 g daily or placebo. Life expectancy was prolonged by approximately a third (median survival 450 and 310 days respectively). In a second study90 of 64 patients with colorectal cancer initially randomised, with a mean follow up of 31 months survival was 96% and 100% (n=27 and n=7 respectively) in treated patients with colon and rectal cancer, and 68% and 53% in the controls (n=19 and n=11 respectively).

In a third small trial91 seven days of cimetidine treatment at the time of surgery was associated with a greater degree of tumour lymphocyte infiltration at surgery than controls (67% and 24% respectively p<0.01). Three year survival was 93% in cimetidine recipients (n=14) and 59% in controls (n=20).

Treatment has been claimed to improve host immunoreactivity by receptor antagonism of suppressor T cells.91 ,92 It is unclear whether there are equivalent or different immunomodulatory properties of other histamine H2 antagonists. On the one hand, metiamide (an H2 antagonist abandoned because it caused bone marrow suppression) and ranitidine were more effective in preventing growth in an animal tumour model of bowel cancer than cimetidine.94 On the other, immunomodulative effects on peripheral blood mononuclear cells were more obvious with cimetidine than with ranitidine or another H2 antagonist, famotidine.95

Examination of data obtained in a surveillance study of 10 000 users of cimetidine followed for 10 years has shown no significant change in colon cancer death rates.96 A randomised clinical trial of cimetidine in 442 cases of gastric cancer has also shown no prolongation of survival, a result at variance with the earlier Danish study.97 Taken overall, data do not give good evidence to support benefit from H2 antagonists in gut epithelial cancer.

Other compounds

A range of other compounds has been extensively studied in animals or cell systems. They include alphadifluoromethylornithine, the dithol ethionine oltipraz, selenium and organoselenium compounds, the adrenal steroid dehydroepiandrosterone, novel retinoids, and flavonoids.

Alphadifluoromethylornithine interferes98 ,99 with putrescine formation by acting as an ornithine decarboxylase inhibitor and there is evidence that it modulates ras proto-oncogene formation.100 Clinically it causes dose related ototoxicity which is reversible, but which limits doses to below 0.5 g/m2.101 Given the long period for which chemopreventative treatment would have to be given, and the wide exposure of individuals who would never develop cancer, the risks may be unduly high.

Retinoids have important prodifferentiating properties but the risks of adverse effects are high because of non-selective actions. The dithiol ethionine oltipraz shows good evidence of chemopreventative properties in animals, the likely basis being increased glutathione transferase activity in tissues, leading to enhanced detoxification pathways.102 ,103 Dehydroepiandrosterone also reduces experimental animal tumour induction possibly through a non-specific adrenal steroid action, or possibly in some way associated with modulation of sex steroid receptors in the bowel epithelium.104

Reduced frequencies of bowel cancer or adenomata have been noted in those with relatively high but still normal blood or tissue selenium concentrations,105 and organoselenium compounds seem to be protective in animals.106 However, others have found no correlations between exposure and protection.107 ,108 The case for flavonoids is limited110 but interesting, as is that for other compounds such as resveratrol which is derived from red grapes.111


There are many indications of possible therapeutic benefits for a range of (at the present time mainly) vitamins and for anti-inflammatory agents. Data are undoubtedly strongest for the anti-inflammatory group.

There is a large body of epidemiological evidence suggesting that aspirin and non-aspirin NSAID use is associated with reduced risks of adenomatous polyp development and of cancer. In addition polyp formation in familial adenomatous disease, but perhaps not in sporadic disease, seems to be retarded by treatment with sulindac. Cancer induction by experimental carcinogens is also inhibited, cell growth rates are reduced and apoptosis induced. Although effects may be associated with COX, particularly COX-2 inhibition, there are suggestions that the mechanisms may be more complex because apoptosis can be induced by related substances which are not prominent COX inhibitors. Aspirin and the other NSAIDs are of particular interest because their actions seem to include prevention of cancer or adenoma occurrence, and interference with the behaviour of neoplastic tissues in vivo or in vitro.

If aspirin is used then doses of at least 325 mg daily are likely to be needed, with full doses of non-aspirin NSAIDs as alternatives.

Trials of anti-oxidant vitamin regimens based on vitamins A, C and E have given inconsistent results and such a regimen now seems unpromising. Use of retinoids alone will have to be viewed with caution in view of evidence that survival in lung cancer is adversely affected, and because at least one retinoid seems to interfere with beneficial actions of vitamin D in retarding cancer cell growth.

Vitamin D may be the next most promising substance for trial after the NSAIDs. There is epidemiological evidence suggesting that exposure to sunlight and raised serum concentrations reduce disease risk. There are experimental data to show that there are functional vitamin D receptors in human colon cancer cells, and growth of colon cancer cell lines appears to be inhibited by OH D3. Further exploration of these properties, which, like those of NSAIDs, seem to include prevention of disease occurrence and the potential to alter established disease behaviour, may have been inhibited by the risks of hypercalcaemia, although the epidemiological data relate to non-hypercalcaemic levels of OH D3 and the experimental data have been obtained with very low concentrations of OH D3. Use of new analogues which do not seem to cause hypercalcaemia is unlikely to be justified in preventative or adjuvant settings until side effect profiles are fully characterised and safety in long term use confirmed. Calcium salts probably offer no significant advantages.

Although there are other candidates, supportive evidence is generally limited. Folate is non-toxic, but supportive evidence to justify large scale trials is inadequate. Ursodeoxycholic acid shows promise of value in disease prevention, but the appropriate dose is unclear. N-acetylcysteine is being included in regimens for head and neck cancer, and the outcome of these may help determine value. Histamine H2 antagonists have some limited but inconsistent evidence to favour their use.

Perhaps the most obvious chemopreventative method, at least in women, is that of oestrogen replacement. It has been known for many years that the sex ratio of colon cancer varies with age. Functional oestrogen receptors are present in the bowel epithelium112and cancer risk is notably reduced in patients on oestrogen replacement therapy.113 ,114

General population intervention

There is currently no chemical intervention for colorectal cancer where sufficient evidence has been obtained in randomised clinical trials; on these classic grounds of evidence-based medicine it is not possible to advocate use of any compound for chemoprevention in a general population study. Furthermore, the annualised frequency of cancer or polyps is too low to make a randomised trial in unselected subjects an attractive proposition. Thus, suppose in an average European country at, say, the age of 55 the frequency of new cancers is of the order of 50 per 100 000 per year. Then the randomisation of 200 000 subjects with an agent in use with a putative reduction of risk of 30% would yield at the end of five years 250 cases in the control group of 100 000, and 180 cases in the test group. The difference is likely to be considerably smaller because of failures of compliance or outright dropout, and because a 30% reduction in risk would be a fairly large expectation. A further likely reduction of power would stem from the tendency to recruit individuals who, by reason of lifestyle, were already likely to be at reduced risk. It is important to bear in mind that the effective sample size drops off as the square of the non-compliance rate in the treatment group and the square of the self-medication rate among the control group the effects of these on the statistical power of a trial can be dramatic.115

It is well known that individuals with adenomata, particularly large adenomata, are at increased risk of second or subsequent polyps. Selective recruitment from such a population would have intrinsic attractions. Firstly, subsequent disease risk is raised and therefore the numbers requiring recruitment would be reduced. Secondly, the chances of patients adhering to treatment schedules would be likely to be raised because of their primary disease profiles. Data obtained in such a study would in essence be concentrated upon rates of adenoma development, with little emphasis upon cancer occurrence, and not upon cancer behaviour since few invasive cancers would be likely to be detected. Nevertheless, such a study has obvious attractions: colorectal polyps present one of the most useful intermediate end points (biomarkers) at the present time.


Patients who have successful resections of primary colorectal neoplasms are known to be at risk of second, synchronous or metachronous cancers, and of synchronous or metachronous adenomata. Studies of these patients would therefore be able to consider polyp prevention, second cancer prevention, and prevention or delay of metastasis or suture line recurrence. Patients who have had apparently successful resections are likely to be well motivated to embrace preventive measures, and would therefore be expected to have high compliance rates. For all these reasons a combined study of adjuvant therapy and adjuvant chemoprevention has obvious attractions.

Data obtainable should include comparative information on mortality patterns—all causes and colon cancer; the occurrence of metachronous polyps, as all adenomata and as pathologically advanced lesions; the occurrence of metachronous colorectal cancer; and on co-morbidity.


The overall five year survival of patients with colorectal cancer is of the order of 50–55%, with deaths in the main being due to cancer spread. In advanced colorectal cancer, in which curative resection is possible, five year survival in Dukes’ B is 45% and drops to 30% in Dukes’ C. Five year survival in resected Dukes’ A is around 80% and survival following simple resection of an adenomatous pedunculated polyp containing carcinoma in situ (or severe dysplasia) or intramucosal carcinoma is generally close to 100%.116Prospects of adequate follow up would be improved if an age cut off, say, of 75 years is applied.

All other causes of mortality will depend upon the age range recruited, but the chances of showing effects are small, except within overall cardiovascular mortality.


The occurrence rate can be estimated by taking the colonoscopic findings in investigations where polyp clearance has been considered complete at colonoscopy. Subsequent polyp occurrence rates in those who have had a primary large bowel cancer will probably be somewhat higher than in patients without initial overt cancer so that figures at the least, will not be likely to overestimate. Data from the United States National Polyp Study suggest a 3.3% rate over a three year period for all adenomata.117 That figure is reasonably close to the 3.6% annual rate recorded over five years by Greenberg et al in a trial of antioxidant vitamins.50


The rate has been estimated at 0.35% per year overall.118 Plainly numbers will be limited in the early years after resection of another primary cancer on the assumption that surgeons will routinely seek second primary tumours and adenomata at the time of initial curative surgery. Furthermore, the rate will be likely to be lower still in patients in surveillance programmes, as will be likely in those with resected cancers.


Breast and ovarian cancer have tended to occur in association with colorectal cancer within families, and there is evidence for common environmental factors. Although ovarian cancer is rare, breast cancer is common, and is roughly twice as common as colorectal cancer. A large preventative study on intestinal cancer might therefore be expected to accrue data on breast cancer. However, assuming a breast cancer incidence rate of 80 per 100 000 per year then in a 1000 patient study of colorectal cancer in men and women the expected disease frequency would still be 0.4 cases per year or less than three cases over a six year period.


Table 3 shows some possible findings over a six year period. It can be seen that a secondary prevention study allows examination of two major problems—the outlook for the initial disease and the prospects of preventing precursor lesions (adenomatous polyps). Treatment might also affect the overall chances of other disease but (outside cardiovascular disease) would be unlikely to show material differences for individual causes. Note that any substantial increase in expected death rates from colonic cancer would increase the chances of detecting treatment effects there, whereas any fall in death rates, overall and/or from cancer would increase the size of the group which could develop new adenomatous polyps. Table 3 emphasises the value of recruiting cancer cases of relatively good prognosis if enough are to survive to provide an evaluable cohort for the study of polyps of a pathologically advanced nature.

Table 3

Possible outcomes in terms of death from recurrent cancer and of diagnoses of new colonic polyps in 1000 patients having successful (clinically curative) colonic resections

Study of patients with treated colorectal cancer has attractions though the potential to examine both an intermediate biomarker, development of new adenomata, and the occurrence of malignant spread from treated disease. Depending on the choice of agent and timing of its use there should be no interference with curative or adjuvant chemotherapy of the conventional type. Studies would also be conducted in patients with a high motivation to comply by reason of treated disease of the same type, and in a group where the risk of further malignancy of the same type, or of recurrent polyps, would be relatively high.

We conclude that there are good reasons for believing that the time is right for the conduct of randomised trials of relatively simple and comparatively non-toxic treatments with biological plausibility in colorectal cancer.


Support from the Associazione Italiana per lo Ricerca sul cancer, the Medical Research Council of the United Kingdom and the Special Trustees of the Former United Birmingham Hospitals is gratefully acknowledged.


Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.