Fortnightly review: screening for asymptomatic colorectal cancer

Inherited colorectal cancer syndromes

The risk of developing colorectal cancer is closely related to a positive family history (table 1)). At the upper end of the risk spectrum lies the dominantly inherited familial adenomatous polyposis syndrome, comprising less than 0.5% of all colorectal cancers. Mutations in the adenomatous polyposis coli (APC) gene are responsible for the familial adenomatous polyposis syndrome (table 2)) and result in hundreds or thousands of colorectal adenomas developing during adolescence and adulthood, with an almost certain risk of adenocarcinoma by middle age.

Hereditary non-polyposis colorectal cancer accounts for 5-10% of all colorectal cancers. Whereas familial adenomatous polyposis results from mutation within a single gene, hereditary non-polyposis colorectal cancer results from a dominantly inherited alteration within one of four DNA mismatch repair genes that have been identified to date (table 2)), which in turn leads to widespread genomic instability. The clinical hereditary non-polyposis colorectal cancer syndrome is defined by the “Amsterdam criteria,” which require the presence of colorectal cancer in at least three family members spanning two generations, with one or more cases diagnosed before the age of 50 years.3 Tumours tend to occur in the right colon, and subjects with the hereditary non-polyposis colorectal cancer (Lynch type II) syndrome also have an increased incidence of gastrointestinal, urinary tract, and gynaecological malignancies.4

Criteria for screening high risk populations

Screening subjects at very high risk (1 in 2 risk) of cancer is the least controversial aspect in the ongoing debate about screening for colorectal cancer. Endoscopic screening is time consuming, expensive, and potentially hazardous but is justified in such subjects because of its sensitivity and specificity for neoplasia. Regular sigmoidoscopy starting in adolescence is indicated for people in families with adenomatous polyposis. Intermittent gastroduodenoscopy with a side or oblique viewing endoscope is also justified in these cases5 because of the high incidence of upper gastrointestinal malignancy in familial adenomatous polyposis.6

The predominance of right sided colorectal cancer in subjects with hereditary non-polyposis colorectal cancer syndrome makes total colonoscopy the endoscopic investigation of choice for such patients, and regular colonoscopy reduces the incidence of cancer in affected families.7 The average age of such patients when cancer is diagnosed is about 44 years,4 7 and screening should begin at 25 years or at least five years earlier than the earliest onset of colorectal cancer in the family.4 8 The optimum screening interval is more contentious: colonoscopy every one to three years is advocated, depending on the presence of neoplasia at initial endoscopy.4 7 8 9 However, Vasen et al report a high percentage of cancers presenting between screening procedures (interval cancers) and propose biennial or annual screening for known gene carriers.10 As with subjects with familial adenomatous polyposis syndrome, subjects with the hereditary non-polyposis colorectal cancer syndrome have an increased incidence of extracolonic cancer, and Lynch et al are investigating the feasibility of screening for gynaecological malignancies in affected families.4 Gastroscopy, urinary cytology, and mammography have also been advocated for this group,7 11 and table 3) shows the recommendations adopted at the 1996 international collaborative group meeting on hereditary non-polyposis colorectal cancer.12

A carefully constructed family history is crucial to determine a person's risk of developing colorectal cancer (table 1)).13 However, identifying the point at which the benefits of endoscopic or radiological screening are outweighed by their disadvantages is perhaps the most difficult task for those with an interest in screening high risk populations. Most experts would offer such screening to subjects who have two or more first degree relatives with colorectal cancer,11 13 14 15 but what about subjects with only one first degree relative affected? Screening recommendations in these subjects have been based on estimates of risk rather than a documented decrease in mortality from colorectal cancer after intervention. Fuchs et al calculated that subjects with an affected family member had a relative risk 1.7 times greater than those with a negative family history16 and supported recommendations that patients with a positive family history should undergo colonoscopic screening from the age of 40,14 especially if the affected family member was under 55 years old at diagnosis.17

Few studies have actually examined the feasibility of screening subjects with a positive family history. Houlston et al and Carpenter et al provided genetic counselling for relatives of patients with colorectal cancer and offered colonoscopic screening to those with a lifetime risk of 1 in 10 or greater.11 15 Although neither detected any cancer in subjects whose risk was less than 1 in 2, both reported compliance rates of about 90% and acknowledged the beneficial psychological effect gained from counselling and, no doubt, from a normal colonoscopy. If detection of early cancers and polyps was the only positive screening end point, data from small family studies11 15 11 would tentatively support a policy of regular colonoscopic screening only in subjects with a lifetime risk of 1 in 2. However, the psychological benefits of counselling and screening that accrue to those with an affected first degree relative may be important. Until firm data on efficacy become available, it therefore seems reasonable to select a screening policy for each high risk individual after an analysis of the proband's age, possible side effects of screening, and the needs and anxieties of the subject.

Future of screening high risk subjects

Molecular biologists are rapidly unravelling the mysteries of colorectal carcinogenesis, allowing clinicians a more scientific approach to identifying individual subjects in families with a history of colorectal cancer who are at risk. The discovery that familial adenomatous polyposis and hereditary non-polyposis colorectal cancer syndromes are associated with germ line mutations enables the carrier status of relations to be determined via genetic screening. This will provide reassurance for those without mutations and allow finite endoscopy resources to be focused on family members carrying mutant genes.

Genetic testing may also be valuable before surgery for some familial adenomatous polyposis patients. Vasen et al found that the risk of recurrent disease after ileorectal anastomosis was particularly great in patients with mutations after codon 1250 of the APC gene, and they recommended a more radical restorative proctocolectomy for these cases.19

Faecal occult blood testing

Traditional faecal occult blood tests detect the peroxidase-like activity of haemoglobin in the stools and rely on the tendency of colorectal cancers to bleed. Sensitivity is limited because many cancers and polyps bleed intermittently, so that blood is unevenly distributed in faeces. Occult blood tests fail to detect 20-50% of cancers and up to 80% of polyps. Specificity is also low and depends on whether patients avoid dietary sources of haemoglobin and myoglobin.21 Red and white meats, fish, some raw vegetables, and fruits containing peroxidase may yield false positive tests. Sensitivity is increased by rehydrating the slide with a drop of water before testing, but this is at the expense of decreased specificity. A positive test requires further evaluation by colonoscopy or sigmoidoscopy and barium enema.

Four randomised controlled trials are currently examining faecal occult blood testing within average risk populations (table 4)).22 23 24 25 Mandel et al of the Minnesota group were the first to publish mortality figures and found a 33% reduction in colorectal cancer related mortality (from 0.9% to 0.6%) in patients randomised to annual screening.24 In an accompanying editorial Winawer stated, “We now have an effective screening method, and I believe we should use it,”26 yet the Minnesota study raised as many questions about the value of occult blood testing as it answered. The incidence of colorectal cancer was unchanged in the screened population over the 13 year study, and the results were criticised because a high proportion (9.8%) of rehydrated slides were positive. This lack of specificity resulted in almost 40% of annually screened patients undergoing colonoscopy during the study, and it seems reasonable to suggest that the reduction in cancer related mortality arose as much from the large number of colonoscopies performed as it did from occult blood testing per se.27 Further follow up from the Minnesota study will determine whether mortality decreases after biennial testing.

More recently, Hardcastle et al reported the results of the largest occult blood screening programme–over 150 000 subjects followed up for a median of 7.8 years.25 Various methodological differences are apparent between Hardcastle et al's and Mandel et al's studies. In contrast with the Minnesota trial, subjects in the Nottingham study underwent biennial rather than annual occult blood testing, slides were not rehydrated before testing, and some patients with weakly positive tests underwent immediate retesting to reduce the number of false positive results. This resulted in a test positivity of only 2.1% for initial screening and 1.2% for subsequent rescreening, while only 4% of subjects required colonoscopic or radiological investigation. Overall, there were 60 fewer deaths related to colorectal cancer in the test group compared with the controls, a 15% reduction in colorectal cancer related mortality. Interestingly, the percentage of stage D cancers was almost identical in test and control groups (22% v 21%), and the reduction in mortality seems to have been due to a shift from regional (stage C) to local (stage A) disease. A negative aspect in this study was the high rate of interval cancers (28%) compared with screen detected cancers (26%) in the test group. This arose partially because of the relatively low sensitivity of occult blood tests for cancer but also because the authors considered that any cancer arising after a negative test was a potentially “missed” cancer, even if diagnosed more than two years after a negative screen.

Mortality figures from Kronborg et al's Danish trial23 are remarkably similar to those of Hardcastleet al (see table 4)). Kronborg et al found that 22% of cancers in the screened group were stage A and that the proportion of incurable tumours was similar in test and control groups. Mortality from colorectal cancer was reduced by 18% in the screened group, again because of a shift from locoregional to local disease. Further follow up of the Minnesota, Nottingham, and Danish study populations will determine whether the incidence of cancer eventually decreases in screened populations secondary to detection and removal of adenomas.

Mortality results from the Swedish study will soon be available and may help clarify the value of population screening with occult blood tests.22 Interim results showed that 36% of screen detected cancers were confined to the bowel wall compared with 9% of those in controls, and 9% were incurable at diagnosis (25% in controls). This provides further evidence that occult blood testing can detect early colorectal cancers, although the shift to early disease was again diluted by a high proportion of incurable cancers (28%) in non-compliant test subjects and in those presenting with interval cancers.

Flexible sigmoidoscopy

Flexible sigmoidoscopy is both sensitive and specific for distal cancers and polyps while allowing polypectomy (fig 2) and pinch biopsy at the time of initial examination. However, it is also a mildly uncomfortable procedure, entailing a hospital visit and preparatory enema. In addition, 30 to 40% of cancers are beyond reach of a 60 cm sigmoidoscope. Case-control studies show that screening sigmoidoscopy and polypectomy reduces colorectal cancer incidence and mortality due to distal disease in average risk populations.28 29 30 Selby et al reported that the risk of death from distal colorectal cancer was 60% lower in subjects who had undergone rigid sigmoidoscopy compared with controls,28 and was 80% lower in subjects over a 10 year period following flexible sigmoidoscopy.29 These results appear impressive; however, case-control studies have many inherent methodological inadequacies,31 and well constructed randomised controlled trials are necessary to confirm the efficacy of this form of screening.

Fig 2

Endoscopic polypectomy: tubulovillous adenoma viewed through colonoscope (left); snare inserted through colonoscope and manoeuvred into position, encircling the stalk (centre); after electrocautery, transection margin is seen in foreground with polyp in background (right)

• Download figure
• Open in new tab
• Download powerpoint

The American Cancer Society recommend sigmoidoscopy every three to five years for people with average risk and a negative initial examination.17 However, Rex et al found no cancers or large (>1 cm) or dysplastic polyps in any patient examined a mean of 3.4 years after a normal flexible sigmoidoscopy,32 suggesting that a longer screening interval is more appropriate. This is supported by data from Aitkenet al showing that subsequent cancer formation is a rare early event, even in patients who have had polyps removed at initial examination.30

Aitken et al have extrapolated the data on endoscopic screening and have proposed a single flexible sigmoidoscopy between the ages of 55 and 60 as a potentially effective screening strategy.33 Their rationale is that the sequence of changes from polyp to cancer is a slow and orderly process taking perhaps 10-25 years to complete. Since colorectal cancer is often diagnosed in the seventh decade, sigmoidoscopy of people in their late 50s would be expected to identify pathologically early distal tumours and precancerous lesions in a proportion of cases. In addition, sigmoidoscopic screening might also indirectly identify a substantial number of proximal neoplasms because large or villiform polyps in the distal large bowel serve as markers of susceptibility for proximal disease.30 Subsequent colonoscopic screening of this relatively small group of patients (estimated at 3-5%) with important distal adenomas might be expected to identify up to a third of proximal adenomas or cancers.33

Aitken et al assume a 70% compliance rate within the population for a single sigmoidoscopic screen, and suggest that sigmoidoscopy and polypectomy (with subsequent colonoscopy in a proportion of patients) might prevent 5500 colorectal cancers a year in the United Kingdom. Although it would take almost 200 000 subjects (65 000 of whom would undergo sigmoidoscopy) and 10-15 years to show whether a survival benefit for this form of screening exists,33 a controlled trial of single flexible sigmoidoscopy might answer important questions relating to endoscopic screening. Indeed, such a multicentre study has started recently, and preliminary data suggest that single sigmoidoscopy is both logistically feasible and acceptable to patients.34

Screening costs

The cost of initial occult blood testing and subsequent colonic assessment is largely unknown. The Nottingham group estimate that about 20% of costs might be recouped by removing polyps that would otherwise progress to symptomatic cancers requiring surgery.35 Mathematical models have also been constructed to determine screening costs,13 36 13 and it is estimated that faecal occult blood testing and subsequent colonic evaluation of those with positive results would cost the United States at least $1.2 billion a year at 1991 costs,36 rising to $2.5bn by the year 2000.38 Regular sigmoidoscopy starting at 50 years of age is even more expensive than faecal occult blood testing and would increase American screening costs by over $5bn a year.13

Although the above figures suggest that population screening would be expensive, they tell us little about the costs in relation to the benefits that might ensue from screening. Preliminary estimates from Britain suggest that a single flexible sigmoidoscopy between 55 and 60 years might be a relatively cost effective screening approach, with an anticipated annual expenditure of £30m or £8500 per cancer death prevented (at 1993 costs).33

More recently, Lieberman constructed a comprehensive cost-benefit algorithm which incorporated not only initial screening expenses but also the subsequent cost of treatment, complications, and additional endoscopic follow up of patients who had polyps removed at first endoscopy.37 With 100% compliance at all stages, different screening regimens were estimated to cost between $225 000 and $280 000 per death prevented, rising to about $350 000 as compliance fell to a more realistic 50%. Compliance and the cost of colonoscopic evaluation were the keys to cost effective screening in Lieberman's study, but all cost-benefit analyses will remain provisional without reliable data about sensitivity, specificity, compliance, and mortality for different forms of screening.

Screening compliance

Lieberman and Hardcastle et al have noted that compliance is an important aspect of screening populations at average risk.37 39 Acceptable initial uptake rates (53-67%) were achieved in all three European trials of faecal occult blood testing, in which subjects were randomly recruited with mailed invitations and reminder letters.22 23 22 Compliance was somewhat higher (75%) in the Minnesota study,24 probably because participants were recruited from members of the American Cancer Society and other highly motivated groups. However, uptake rates of under 25% are often observed in uncontrolled trials.40 Furthermore, in Germany faecal occult blood testing has been offered nationally since 1977, but only 21% of women and 10% of men take advantage of such screening.41 Satisfactory compliance levels have occasionally been observed for flexible sigmoidoscopic screening programmes, with acceptance rates as high as 70%.42 In uncontrolled studies of screening sigmoidoscopy, however, levels rarely rise above 30%.43

Compliance in average risk populations depends on demographic, behavioural, and educational factors.44 Many people in the general population are ignorant of the sequence of adenoma to carcinoma, the concept of asymptomatic disease, and the potential benefits of screening. Some non-responders also suppose that they are less susceptible to cancer than others,45 which is clearly erroneous since the incidence of colorectal cancer in non-responders from the initial Nottingham trial was 0.92 per 1000 person years compared with 0.72 per 1000 in the controls.39 These issues might be addressed by public health education focusing on the concept of asymptomatic and premalignant disease and the benefits of early detection.45 Education could also highlight the high incidence and hereditary aspects of colorectal cancer in Western society and the aetiological role of a high fat, low fibre diet in carcinogenesis.

Although compliance is relevant to the success of population screening, it remains questionable as to how far healthcare agencies should go to promote screening to potentially unreceptive populations.46 Statements such as “Mammography helps your doctor see breast cancer before there is a lump when the cure rates are near 100%”47 and “Given the overwhelming evidence that (colorectal cancer) screening is effective in detecting and curing this second deadliest cancer”48 may increase compliance, but they give an inaccurate perception of the efficacy of screening. Public health education might also reasonably include information about the emotional costs of false positive tests, the inappropriate reassurance caused by false negative results, and the possibility of detecting incurable disease. This would allow people to make informed decisions about participating in screening programmes.

Compliance might also be improved by tackling concerns about the specific type of screening offered and its mode of delivery. A proportion of non-responders find the concept of faecal occult blood testing aesthetically unacceptable,49 while dietary restrictions before screening also adversely affect uptake rates.50 Strategies to overcome people's distaste for screening could include developing more acceptable, and possibly self administered, tests.49 Participation rates would also be increased by involving well motivated primary healthcare staff in the delivery process.51

Criteria for screening average risk populations

There is much debate about the desirability or need for a colorectal cancer screening programme for the general population. The American Cancer Society recommend an annual digital rectal examination from the age of 40, annual faecal occult blood testing from age 50, and sigmoidoscopy every three to five years from age 50.17 Faecal occult blood testing is offered in Germany as part of an annual cancer checkup.52 In contrast, the King's Fund and the Canadian Task Force on the Periodic Health Examination do not currently advocate screening in average risk subjects.53 54

What should doctors make of these conflicting messages? It is clear that intensive screening by competent researchers can reduce mortality from colorectal cancer, but it is questionable whether this in itself justifies establishing a national screening programme. Overall, we believe it is premature to implement such a programme until there are reliable data on the acceptability, financial costs, logistics, and the potential physical and emotional side effects of screening. In this context it is useful to review the results of long established breast and cervical cancer screening programmes that have recently been criticised as hugely expensive and largely ineffective.46 55 56 57 In one of the largest studies of its kind, Raffle et al concluded that “The real lesson from 30 years' cervical screening is that no matter how obvious the predicted benefit may seem from any screening test, introduction should never take place without adequate prior evaluation of both positive and negative effects in controlled trials.”57

When applied to colorectal cancer, the overall benefits of conventional screening do not seem particularly obvious. Screening by a single sigmoidoscopy is no more than an attractive concept, while regular sigmoidoscopy or faecal occult blood testing seems excessively expensive at present. In addition, the high rate of false positive faecal occult blood tests would cause substantial distress to many healthy subjects where none would have existed but for screening.

Future of population screening

Current results are encouraging enough to warrant further research into more effective screening regimens for average risk subjects. Increased cost effectiveness and more refined screening strategies may eventually tip the balance in favour of population screening. The cost effectiveness of sigmoidoscopic screening would increase considerably if gastroenterologists were replaced by nurse endoscopists for diagnostic examinations. Large studies have concluded that nurse practitioners can perform flexible sigmoidoscopy as accurately and safely as trained gastroenterologists or colorectal surgeons,58 59 and patients' pain, bloating, or embarrassment is no greater when examinations are performed by nurses rather than gastroenterologists.58

The cost effectiveness of faecal occult blood testing may also improve with the introduction of new tests. Haemoccult II Sensa (SmithKline Diagnostics) is a modification of the Haemoccult II test, while Haemselect (SmithKline Diagnostics) is an immunoassay for human haemoglobin. Both have higher sensitivity than Haemoccult for blood loss in symptomatic60 and asymptomatic patients,61 62 and, when used in combination, their specificity is close to that of unrehydrated Haemoccult II.62 Increased sensitivity without a corresponding loss of specificity might substantially improve the performance of faecal occult blood testing63 and eventually result in relatively cost effective screening.

Molecular biology may also be used in the future. Genetic mutations are found with relative ease in DNA extracted from tumour tissue, but they can also be detected in DNA recovered from the sputum, urine, pancreatic juice, and faeces of patients with various different malignancies. Sidransky et al and Smith Raven et al recently isolated K-ras mutations in faeces and suggested that genetic analyses might eventually evolve into screening tests for colorectal cancer.64 65 There are certain advantages in screening for gene alterations rather than occult blood in stool. DNA is extremely stable, whereas blood, especially from the right colon, may be degraded by bacteria and subsequently result in false negative faecal occult blood tests. Mutations are also highly specific for neoplasia, virtually eliminating the problem of false positive results.

The range of somatic mutations that can be detected in faecal DNA64 65 66 is too limited to allow practical genetic screening studies at present, and many molecular biological assays are unsuitable for routine clinical use. However, if developments in molecular biology continue to unfold at their current pace the concept of a non-invasive, sensitive, and specific genetic screening test for colorectal cancer may not be as far fetched as it seems.