In this paper, we do not assess the decision of whether to screen for colorectal cancer; that topic has been discussed elsewhere ([13, 19, 20]; Eddy DM. Screening for colorectal cancer. In preparation). Instead, we focus on the interpretation and follow-up of results of screening with fecal occult blood tests. When such screening is done, clinicians confront several specific issues about technique. These issues may seem mundane, but they can greatly affect on the effort (including cost) and efficacy of screening. Among the issues addressed here are the procedures for screening with fecal occult blood tests, the definition of a positive test result, the timely evaluation of positive results, and the rescreening of persons with positive or negative results. Authoritative guidelines about screening for colorectal cancer have been remarkably silent on some of these important details.

1. Methods

We identified relevant studies by searching the MEDLINE database for articles published between 1984 and 1996, using the terms fecal occult blood and colorectal cancer screening, and we reviewed articles from the bibliographies of materials identified through the MEDLINE search. We also searched for studies about the biology of colorectal cancer as it relates to screening with fecal occult blood tests, the validation of fecal occult blood tests, the clinical application of screening with fecal occult blood tests, and the cost-effectiveness of screening for colorectal cancer. This review is not a structured meta-analysis on trials of screening with fecal occult blood tests or of the validation of such tests because of the paucity of published data that meet typical criteria for assessing validation. However, we sought all published data that might relate to the optimal performance of fecal occult blood tests. The strength of estimates for each measure reflects the authors' overall assessment of the evidence and was graded subjectively because of the need to consider several types of studies. The special methodological challenges in assessing screening with occult blood tests are identified and emphasized in the body of this report. This discussion and comments in the guidelines published in this issue [21] provide details so that readers can judge the nature and quality of the evidence on their own.

2. Data Synthesis

2.1 Reducing Mortality Rates Associated with Colorectal Cancer by Screening with Fecal Occult Blood Tests

Three clinical trials have credited screening done using fecal occult blood tests with reducing the mortality rate associated with colorectal cancer. A trial in the United States studied volunteers who were screened using an occult blood test with rehydrated slides. In the group of persons who were screened annually, mortality rates associated with colorectal cancer were reduced by about 33%; 39 deaths from colorectal cancer were prevented among 15 570 persons during the 13 years of screening. These data translate to a relative risk reduction for death from colorectal cancer of 33%, an absolute risk reduction of about 2.95 deaths per 1000 persons, and a number needed to screen to prevent one death over 13 years of about 300. Two European trials that reported mortality rates were population based and shorter in duration; these trials used biennial screening and no rehydration of slides. They reported a reduction in mortality rate of 15% to 18% [3, 4]. A third European trial is still being conducted [22]. Because compliance was about 50% in these trials, screening efficacy may be higher than reported effectiveness. In any case, it is now clear that screening with fecal occult blood tests can reduce the mortality rates associated with colorectal cancer. The major challenge now is to determine how well such screening works and how it can be improved. These randomized clinical trials provide important data not only about reduction in mortality rates but also about the ways in which clinicians and patients can interpret the results of fecal occult blood tests.

Table 1 clearly shows that no consensus exists (at least as reflected in the design of the clinical trials) about such important technical issues as diet restriction, rehydration of test slides, frequency of screening, and the strategy for evaluating persons who have positive results (that is, retesting patients with positive results before doing a work-up or immediately doing a work-up for all patients with positive results). Because the trials also differ with regard to such features as targets of study (volunteers or population based), compliance, and duration of follow-up, it may be difficult to sort out which features have the greatest effect on mortality rates. Table 2 summarizes some of the expected results on screening with fecal occult blood tests based on data from randomized clinical trials.

Although it is now agreed that mortality rates associated with colorectal cancer can be reduced by screening with fecal occult blood tests, the degree of reduction in these rates and the approaches needed to decrease them further remain uncertain. The degree of reduction in mortality rates ultimately depends on the sensitivity of fecal occult blood tests; thus, because false-negative results do occur, mortality rates associated with colorectal cancer will not be optimally reduced by screening with occult blood tests. In addition, patients and physicians must not interpret a negative test result to mean that the patient is protected from colorectal cancer. A corollary is that worrisome signs and symptoms that occur after a negative screening result must still be assessed vigorously. Clinicians can also anticipate the development of new strategies to improve screening efficacy by improving test sensitivity or specificity.

2.2 Types of Fecal Occult Blood Tests and Procedures for Performing Them

All screening with fecal occult blood tests is based on the concept that important target colonic neoplasms, such as early-stage cancer (Dukes stage A or B) and large adenomatous polyps, will bleed and may be detected by an occult blood test. When implementing screening with fecal occult blood tests, clinicians must choose from among various types of tests and techniques for applying these tests. The types of occult blood tests that are currently available and the procedures used to perform the tests are discussed below.

2.2.1 Types of Tests

The fecal occult blood tests currently available in the United States include such guaiac-based tests as Hemoccult II and HemoccultSENSA (SmithKline Diagnostics, San Jose, California), such immunochemical tests as HemeSelect (SmithKline Diagnostics), and the HemoQuant test (Mayo Medical Laboratories, Rochester, Minnesota) [28, 29]. When gastrointestinal blood is lost, the stool will contain a combination of intact or nearly intact hemoglobin, intact heme, and heme-derived porphyrins in amounts that depend on the site and amount of bleeding and the transit time through the gut. These blood products are detected in different ways by the various tests. Several reviews [28-32] discuss the mechanics of fecal occult blood testing and nonneoplastic causes for positive results.

Several mechanisms are used to detect occult blood. Guaiac-based tests detect occult blood on the basis of the pseudoperoxidase activity of heme or hemoglobin; such activity converts colorless guaiac to a blue color in the presence of the hydrogen peroxide in the developing reagent. The HemoQuant test detects some heme and heme-derived porphyrin. Immunochemical tests detect intact or nearly intact human hemoglobin by various immunochemical techniques; at least theoretically, this technique is more specific for detecting loss of blood from the lower intestine because blood from lower sites is less degraded during transit.

Because Hemoccult II and HemoccultSENSA are simple tests that can be done in a physician's office, they are considered waived laboratory tests according to the Clinical Laboratory Improvement Amendments of 1988. The HemeSelect test, which is similar to an enzyme-linked immunosorbent assay, is more complex and thus must be done in a laboratory. Although this test could be done in many laboratories, few currently perform it. The manufacturer has developed and is planning to market a simpler immunoassay, FlexSure OBT (SmithKline Diagnostics), which can be done in a physician's office (Baker J. Personal communication). Ease of use could help reduce what may be a substantial barrier to the implementation of immunochemical tests. A product that is identical to HemeSelect is used in mass screening in Japan; this test is processed using automated instrumentation [33].

2.2.2 Characteristics of Various Tests

The sensitivity and specificity of various fecal occult blood tests can be compared across different studies among different patients or in the same study among the same patients; the latter type of comparison provides the more reliable estimates of the relative strengths of various fecal occult blood tests [27], as shown in Table 3. In the study shown, four strategies were compared (rehydration of Hemoccult II was not one of the strategies).

For the data in Table 3, a concurrent structural examination of the colon (for example, colonoscopy) was not used as the gold standard to determine whether colorectal cancer was present. Instead, follow-up for 1 year after screening was used to identify colorectal cancer that may have been present despite a negative result on fecal occult blood testing [27]. In this kind of assessment using follow-up, estimates of sensitivity necessarily decrease over time (for example, over 2 or 4 years) as more cases of colorectal cancer are discovered in patients who had negative test results; these are then counted as false-negative results [34]. Although the rates of absolute sensitivity are difficult to estimate, partly because it is difficult to know whether colorectal cancer is present but undetected, useful information can still be obtained by comparing the relative sensitivity of various fecal occult blood tests applied to the same patients [35]. In such a comparison, the relatively high rate of sensitivity for HemeSelect compared with Hemoccult II (Table 3) is supported in other studies, although those studies used persons with symptomatic cancer [28, 36]. The high rate of false-positive results on HemoccultSENSA seem to make it impractical as a stand-alone test [37, 38].

A new approach has recently been proposed to increase sensitivity and decrease the rate of false-positive results. The idea is to use a two-step approach in which one stool specimen is collected on each of 3 days and samples are prepared for HemoccultSENSA and HemeSelect. The more sensitive HemoccultSENSA test is developed first and, if the result is positive, the more specific HemeSelect test is then developed. The overall results of the screening are considered positive only if the results of both HemoccultSENSA and HemeSelect are positive. If the HemoccultSENSA test results are negative, then the samples that were collected for the HemeSelect test are discarded [27]. This approach may be substantially better than the widely used Hemoccult II test.

Because the HemeSelect test (which is not guaiac based) does not yet seem to be practical, we believe that the fecal occult blood test used for screening in the United States should be a guaiac-based test, such as Hemoccult II. However, the two-step approach [27] is promising and, if it can be made widely available and practical [37, 38], could become the preferred method of screening. An office-based immunologic assay, such as FlexSure OBT, may provide a single-step test with increased sensitivity and specificity.

2.2.3 Procedures for Testing

The currently accepted procedure for the Hemoccult II test involves preparing one slide with two windows each day and sampling two sites from three stool specimens. This procedure was established by historical precedent [39] rather than empirical research; it is based on the concern that colorectal neoplasms may bleed intermittently and the concern that blood may not be homogeneously distributed in a specific bowel movement. Positive screening results have commonly been defined as one or more positive reactions out of the six windows prepared when assessed for a blue-color reaction at 30 to 60 seconds. The goal of using slides collected over several days and interpreting the test result as positive if any window is positive is to maximize the sensitivity of the test for colorectal cancer and precursor neoplasms. Obtaining a fecal sample for an occult blood test during a rectal examination, although commonly done in practice, is not recommended; data are insufficient to allow the interpretation of a positive test result obtained during a rectal examination. One concern is that false-positive results may be induced by trauma from the rectal examination itself.

2.2.3.1 Number of Windows and Definition of Positive Results. The sensitivity and specificity of Hemoccult II may be influenced by the number of samples used and the criterion for how many windows need to yield positive results before the overall test results are considered positive. Even the decision about whether a window has enough blue color to be interpreted as positive has been debated [40, 41]. Although various strategies for Hemoccult II have been considered (for example, using more or fewer slides or different criteria for positivity), the evidence needed to assess such strategies is lacking [42-45]. Until better data are available, the preferred method for Hemoccult II should continue to include the preparation of two-window slides from three separate fecal specimens and the definition of a positive result as one or more positive windows.

2.2.3.2 Diet and Drug Interactions. Diet may affect the results of fecal occult blood tests. For example, false-positive results (positive results that occur when no bleeding is present) on guaiac-based fecal occult blood tests may result from intake of foods that contain peroxidase activity [29]. It is not clear whether low doses of aspirin or even warfarin cause clinically significant blood loss and false-positive results on fecal occult blood tests [46, 47], but higher doses of aspirin may cause problems [48]. False-negative results may be due to intake of vitamin C [49]. Causes of false-positive and false-negative results have been discussed in detail elsewhere [30-32].

In most of the clinical trials (Table 1), persons being screened were encouraged to abstain from red meat, poultry, fish, some raw vegetables, vitamin C, and aspirin [2]. Although compliance with diet restriction has not been reported in detail, recommending a strict regimen may decrease overall compliance with screening [50]. The topic of compliance deserves further research because its effect on test results and subsequent mortality rates may be substantial. Given the data that are currently available, we believe that low-dose aspirin and warfarin may be continued if necessary during fecal occult blood testing and that restrictions of diet and intake of nonsteroidal anti-inflammatory drugs should be encouraged but not overemphasized because of possible detrimental effects on compliance. When a person who did not restrict medication or diet has positive test results, work-up should still be done without retesting after further diet or medication restriction.

2.2.3.3 Rehydration of Samples. Perhaps the most controversial unresolved issue about the method of screening for colorectal cancer with guaiac-based fecal occult blood tests is whether to rehydrate slides by adding a drop of water to the slide window before development with the peroxide-containing reagent. The goal of rehydration is to increase test sensitivity and the rate of positive results, which decline with time from slide preparation to development. In the Minnesota study, the overall test positivity rate increased more than fourfold, from about 2% without slide rehydration to about 10% with rehydration (the range was from 8% in younger patients to 16% in older ones), and reported sensitivity increased from 80% to 92% [2].

The decision about rehydration is controversial because it involves a potential tradeoff between missed cancer if rehydration is not done and an increased number of more extensive work-ups (generally colonoscopy) if rehydration is done. The issue is further complicated because the fraction of the colonoscopic work-ups that are done because of false-positive results on fecal occult blood tests when the neoplasm detected did not bleed enough to cause the positive result [2] may reduce mortality rates when a neoplasm is discovered by chance in a work-up [51, 52]. However, as the rate of positive results increases, the resultant volume, effort, and risks of colonoscopic work-ups that are required to achieve additional benefit will constitute important practical challenges to rehydration (Eddy DM. Screening for colorectal cancer. In preparation).

It is our opinion that a rate of false-positive results of 10% or more (Table 1) is unacceptable, especially in light of the relatively small apparent increase in sensitivity with rehydration. Similar conclusions have been reached in other studies, including an analysis based on quantitative modeling ([13]; Eddy DM. Screening for colorectal cancer. In preparation). Although we are not aware of formal surveys on whether clinicians actually rehydrate Hemoccult II slides, we think that most do not. Combined rehydration and annual testing dramatically increased the rate of colonoscopy to a cumulative 31% for a period of about 13 years in the Minnesota study (Mandel J. Personal communication) compared with rates of about 4% for about 10 years of biennial, nonrehydrated screening with fecal occult blood tests in Denmark and the United Kingdom [3, 4].

2.2.3.4 Timing of Development of Test Samples. Seemingly mundane practical decisions, such as the delay between sampling and development of the samples for fecal occult blood tests, may substantially affect test results and interpretation. For example, after specimen collection, the rate of true-positive results (detection of blood) and the rate of false-positive results (detection of vegetable peroxidase) begin to decrease over time; however, the decrease of true- and false-positive results occurs at different rates because vegetable peroxidases degrade more rapidly (Young G. Personal communication). Because a clinician's goal is to maximize true-positive results and minimize false-positive results, these different rates of decrease become important in considering when (that is, how many days after stool sampling) to develop cards for fecal occult blood tests. Test manufacturers and clinical investigators are largely silent on the issue of timing, and it deserves further investigation.

2.3 Test Characteristics of Fecal Occult Blood Tests

Clinicians who use fecal occult blood tests to screen for colorectal cancer should be aware of the sensitivity, specificity, and predictive value of such testing and of how these indices of test performance have been estimated. These data are important because they have critical implications for the interpretation of positive and negative test results.

The main features of these indices are discussed below, including 1) the sensitivity of a screening test [for early colorectal cancer or for large polyps], which is the most important determinant of benefit; 2) specificity, which determines effort and cost; and 3) the challenges in estimating sensitivity.

2.3.1 Sensitivity Determines Benefit

The primary determinant of the benefit of fecal occult blood testing is the test's sensitivity for detecting early colorectal cancer and its precursors (such as large adenomas) because benefit obviously can occur only if neoplasms are detected. As a corollary, a fecal occult blood test cannot detect neoplasms that do not bleed. Therefore, the bleeding biology of colorectal cancer and precursor neoplasms ultimately determines the upper limit of screening efficacy for the fecal occult blood test. As discussed below, some and perhaps many cases of early colorectal cancer may bleed infrequently or not at all. We speculate that the sensitivity (of a single test) of any fecal occult blood test for curable colorectal cancer is less than 50% and may be as low as 30%. However, repeated testing using a test with low sensitivity could still produce a large benefit if target neoplasms remain at least intermittently detectable for years before they progress to incurable disease. Unfortunately, little direct evidence quantifies the effects of repeated testing. Because the sensitivity rate for fecal occult blood testing is used in simulation models to project estimates of screening efficacy ([9, 19, 20]; Eddy DM. Screening for colorectal cancer. In preparation), it is important to obtain accurate estimates of sensitivity. Moreover, clinicians need to know about sensitivity rates because these rates affect how the result of a screening test may be interpreted for a patient; in particular, sensitivity rates determine how well colorectal cancer can be excluded by negative results on a single screening test or by a program of repeated tests. Some of the methodological challenges involved in deriving these estimates are described below.

2.3.2 Specificity Determines the Cost and Effort of Screening

The specificity of fecal occult blood tests is the principal determinant of the effort of screening because false-positive results account for almost all colonic examinations required in working up results of fecal occult blood tests. Clinicians should understand that seemingly small numerical changes in test specificity may translate into large effects on the total effort of screening, an increase that could occur without a substantial increase in benefit. For example, if specificity decreases from 98% to 96% so that the false-positive rate increases from 2% to 4%, the number of false-positive results doubles. If the false-positive rate increases to 8% or more (as it does with Hemoccult II rehydration), the total rate of positive results increases fourfold. When many persons are screened, many false-positive results occur as a consequence of intake of foods with peroxidase activity, non-neoplastic gastrointestinal bleeding (such as that from hemorrhoids or angiodysplasia), or unknown reasons.

2.3.3 Probability of Finding an Important Colonic Neoplasm after a Positive Result

A major clinical concern of this report is whether a complete colonic evaluation should be done on a person who has a positive result on a fecal occult blood test. The probability of finding an important neoplasm after a positive result depends on several factors, including the testing technique (for example, whether Hemoccult II was rehydrated) and whether the screening is an initial or repeated examination. It also depends on the definition of an important neoplasm, for example, early colorectal cancer (Dukes stage A or B) or large adenomas. Advanced colorectal cancer is arguably less important because therapy for this condition is less effective, and such very early neoplasms as small adenomas are so common that most are probably clinically unimportant.

Table 1 shows the results of fecal occult blood tests (using Hemoccult II) for mostly asymptomatic, middle-aged adults at average risk for colorectal cancer. Several features that affect results are noted below.

2.3.3.1 Symptom Status. Although the effort of a screening program is directed toward asymptomatic persons, persons with minor symptoms may preferentially participate [24]. If the latter persons participate, the probability of finding colorectal cancer or an adenoma after a positive result on fecal occult blood testing is increased.

2.3.3.2 Rehydration. A work-up done for a positive nonrehydrated initial screening was associated with rates of early colorectal cancer of about 7% in the trial from the United Kingdom and 14% in the trial from Denmark; for detection of any colorectal cancer, the rates were higher. In contrast, for rehydrated samples, the probability of finding colorectal cancer in the early stages was about 3% in the trial in Sweden and was not reported for the Minnesota study.

2.3.3.3 Initial Screening Compared with Rescreening. The probability of colorectal cancer after a positive result on fecal occult blood testing is higher for initial screening than for rescreening. This higher probability is presumably caused partly because bleeding neoplasms are preferentially detected and removed during initial screening. The overall rate of positive results may not decrease much at rescreening (Table 1) because most positive results are false.

2.3.3.4 Inclusion of Large Adenomas as Targets of Screening. In addition to early-stage colorectal cancer, large adenomas are important targets for screening. However, few data address the rates of adenomas in persons with positive results on occult blood testing. Data from clinical trials (Table 1) show rates of about 17% to 46% for early colorectal cancer or large adenomas in persons with positive test results. The inclusion of mildly symptomatic persons in screening [24] could substantially affect the interpretation of these rates. In any case, the rates are impressive. Further study or analysis should be done to stratify the rates among truly asymptomatic and mildly symptomatic persons.

2.3.3.5 Summary. Overall, for middle-aged persons having initial screening with nonrehydrated Hemoccult II, the probability of a positive result on fecal occult blood testing is about 1% to 4%. When a work-up is done, the probability of finding early stage colorectal cancer in patients with a positive result is about 5% to 14%; the probability of finding early colorectal cancer or a large adenoma in such patients is unclear but is probably 20% to 40% or perhaps even greater. The possible influence of including persons with mild symptoms has not been assessed in interpreting these rates.

Although the probabilities of early colorectal cancer that have been reported in clinical trials span a wide range (Table 1), in our view, any of those rates is clinically meaningful and sufficiently high to warrant a timely, complete colorectal evaluation when the results of fecal occult blood testing are positive. Exceptions might be made for persons in whom another source of bleeding seems clear or for younger persons who do not have signs, symptoms, or a family history of colorectal cancer and thus have a very low prior probability for colorectal cancer.

2.3.4. Probability of No Important Neoplasm after a Negative Result

Few direct observational data allow estimation of the probability of no colorectal cancer in persons who have negative results on fecal occult blood testing. In a middle-aged person whose initial screening results are negative, the probability of no colorectal cancer is greater than 99%, partly because the prior probability or prevalence of colorectal cancer at that age is so low at baseline (approximately 0.5%). However, the ability of a negative result on occult blood testing to exclude colorectal cancer should not be overestimated because the sensitivity of such testing is limited. A negative result on fecal occult blood testing has a similarly limited ability to exclude large adenomas because its sensitivity in detecting large adenomas is lower than that for detecting early colorectal cancer [53].

2.3.5 Challenges in Measuring Sensitivity

Calculating an absolute sensitivity rate for fecal occult blood testing is difficult because of several unresolved problems. First, directly measuring sensitivity requires evaluating the whole colon in persons with negative or positive results to determine whether neoplasms have been missed. Because it is usually impractical to evaluate the colon in healthy, asymptomatic persons who have negative test results, sensitivity must be estimated indirectly. Second, screening typically involves repeated testing over time; it is not clear how test sensitivity at a single application relates to the sensitivity of a program of repeated testing. Third, it is not yet clear how much and how often bleeding occurs from such important neoplasms as large polyps or early cancer. A detailed consideration of the issues that are involved in assessing sensitivity is beyond the scope of this report. Nevertheless, clinicians should recognize that answers to such issues may substantially affect our approach to screening.

2.4 Evaluation of a Positive Result on Screening with Fecal Occult Blood Testing

After a positive result on fecal occult blood testing, the basic choice is whether to retest (for example, by doing another series of three Hemoccult II cards while further encouraging restrictions of diet and medication) before doing a work-up or to do the work-up immediately. We prefer to do an immediate work-up because of the clinically significant rate of colorectal cancer or a large adenoma after a positive test result (one or more positive windows out of six), as discussed above. The strategy of retesting before a work-up to increase the specificity of the test [23] remains largely unassessed. We believe that this strategy may cause some cases of colorectal cancer to be missed. After a positive result on occult blood testing, an arbitrary but reasonable guideline for timely evaluation is to perform a complete colonic examination within 2 to 3 months. Longer delays create the potential for increased spread of the tumor, increased patient anxiety, and decreased compliance with follow-up testing. Conversely, emergent or urgent colorectal evaluations are not warranted, given the generally favorable biology of colorectal cancer, because they only create stress in the patient being screened.

Options for evaluating a positive result on fecal occult blood testing include complete colonoscopy or flexible sigmoidoscopy and air-contrast barium enema [54]. Colonoscopy allows diagnosis and treatment (for example, adenoma excision) in one step but incurs a risk, although small, of perforation (about 0.8% for polypectomy during the procedure); perforation can also occur during simple diagnostic colonoscopy [55]. Moreover, complications can result from intravenous sedation. The combination of barium enema and flexible sigmoidoscopy offers lower risk and lower cost. One disadvantage of barium enema, however, is that abnormal examination results requiring subsequent colonoscopy occur in approximately 15% or more of patients [56-58]. Moreover, and of great importance in our view, a barium enema may not detect large ( 1 cm) adenomas in about 40% of cases [59]. The false-negative rate is presumably much higher for smaller adenomas, but smaller adenomas are common and have less clinical significance than large adenomas.

In the United States, where colonoscopy is generally safe and accessible, the preferred strategy for evaluating a positive result on an occult blood test is complete colonoscopy. In some situations, flexible sigmoidoscopy and high-quality double-contrast barium enema may be preferred (for example, in a person with extensive pelvic scarring, in a person receiving anticoagulant therapy that cannot be stopped, or in other situations in which a patient's other medical problems might be aggravated by the sedation or effort associated with colonoscopy). Finally, if colonoscopy is incomplete, the clinician has several options, including repeating the examination (for example, if the reason for failure can be rectified), doing a barium enema, or (in rare cases) simply waiting and repeating the screening examination in the future. The variables to be weighed in this decision include the reason why the examination failed; the amount of the colon that was not examined; and the patient's age, life expectancy, ability to tolerate the procedure, and tolerance for uncertainty.

2.5 Rescreening after Negative Results on Complete Colorectal Evaluation

When the results of screening are positive but the results of the complete colorectal work-up are negative for cancer or adenoma, the issue of whether to discontinue screening for a few years arises. Several types of indirect evidence suggest that a negative result on colonoscopy is a powerful predictor for very low short-term risk for colorectal cancer [6, 60], and some direct evidence may soon be available on this point [61]. This issue is quantitatively substantial in screening programs because, for example, about 13% of all diagnostic colonoscopies in the Minnesota study were done in persons who had been reentered in a screening program after having negative results on a colonoscopic work-up within 5 years (Mandel J. Personal communication). The period during which a patient will remain at low risk after negative results on colonoscopy, however, is still uncertain. We believe that after such results, screening with occult blood tests can be deferred for 5 years; it must be understood, however, that some persons may develop colorectal cancer, even during that brief time, for two reasons. First, some neoplasms may be missed by colonoscopy [60]. Second, some cases of colon cancer may grow more quickly than others [62, 63]. In any case, it is clear that negative results on colonoscopy do not totally "protect" the patient against colorectal cancer [6]. Despite these limitations, colonoscopy seems to provide important protection from colorectal cancer and important information about future risk for colorectal cancer.

When results of screening with fecal occult blood tests are negative, screening with fecal occult blood tests should generally be continued in 1 year (or another form of screening could be done). Because persons who have had negative results on fecal occult blood testing are by no means guaranteed freedom from death as a result of colorectal cancer, signs and symptoms that develop between subsequent screening examinations should be appropriately evaluated.

2.6 Surveillance Colonoscopy in Persons Found To Have Colonic Adenomas at Screening

Until recently, authorities recommended that all persons who have had any colorectal adenoma should undergo periodic colonoscopic surveillance for their entire lives [8, 64-66]. The decision about surveillance is clinically and economically important because colorectal adenomas are so common and will therefore be found when screening is done. Details about strategies for surveillance after adenomas are found are beyond the scope of this discussion and have been considered elsewhere; however, it is now generally accepted that when only one small adenoma is found on colonoscopy, no special subsequent surveillance is warranted and persons can be followed with routine screening [10, 17, 67, 68].

2.7 Cost and Cost-Effectiveness of Screening

Both the cost and cost-effectiveness of screening with fecal occult blood tests are relevant to decisions about the implementation of screening ([13, 20]; Eddy DM. Screening for colorectal cancer. In preparation). Several strategies are available for screening for colorectal cancer (for example, fecal occult blood testing alone, sigmoidoscopy alone, sigmoidoscopy and occult blood testing combined, or periodic barium enema) and seem to cost about $15 000 to $25 000 per year of life saved ([9, 20, 69, 70]; Eddy DM. Screening for colorectal cancer. In preparation), and a case can be made in favor of periodic colonoscopy alone [71-74]. The results of different strategies are similar enough that some authorities have chosen not to select between fecal occult blood testing and sigmoidoscopy ([13]; Eddy DM. Screening for colorectal cancer. In preparation). Moreover, these cost-effectiveness estimates seem to be similar to those of other common preventive strategies, such as screening mammography [75]. However, we expect that the magnitude of cost-effectiveness may be substantially affected by such features as the degree of reduction in mortality rates (for example, whether mortality is reduced by 15% [3, 4] or 33% [2]), false-positive results on occult blood testing (Table 1), and patient compliance and age. Simultaneous consideration of these features may help improve the overall screening effort.

Even if screening for colorectal cancer is cost-effective, it is nevertheless costly. Direct cost was therefore a major factor in the decision of the U.S. Congress to not fund screening for patients in the Medicare program, although the analysis [76] done by the Office of Technology Assessment showed that screening was cost-effective (Wagner JL. Personal communication). One major source of screening cost is the colonoscopic work-up of patients who have positive results on fecal occult blood tests. If colonoscopy is done on the basis of false-positive screening results (that is, no early colorectal cancer or large polyps found at work-up), the cost of the procedure leads to minimal or no benefit. Therefore, the reduction of false-positive results is a particularly important consideration in the economics of screening for colorectal cancer.

3. Discussion

3.1 Authoritative Guidelines for Screening for Colorectal Cancer

Although almost universal consensus about whether to screen exists among authoritative organizations, few recommendations are available on such specific technical issues as how to perform the screening test (for example, using rehydration of samples, dietary modification, or restriction of medication) or how to handle subsequent surveillance by colonoscopy that may be prompted by screening results. All of these issues affect the cost and results of screening for colorectal cancer. The American Cancer Society recommendations of 1980 and 1991 [77, 78] offer no guidelines on rehydration and surveillance colonoscopy, although surveillance is discussed in the revisions of 1992 [10]. The guidelines from the U.S. Preventive Services Task Force [13] and those initiated by the Agency for Health Care Policy and Research [19] do not recommend rehydration. None of the five sets of recommendations suggests how or when to implement rescreening if the results of a colonoscopic work-up are negative. A separate concern is the decision about when to start and stop screening, for example, in relation to a person's absolute risk for colorectal cancer and life expectancy. A case can be made for starting screening at older ages, such as 55 or 60 years, rather than at 50 years of age, as is currently recommended [79]. In the future, authoritative guidelines should specifically address these issues.

3.2 Summary and Discussion of Management Principles

In making a decision about screening for colorectal cancer with fecal occult blood tests, the clinician must bear in mind the issues discussed here. Specific recommendations are given elsewhere in this issue [21]. The best current approach to screening with fecal occult blood tests seems to be to use Hemoccult II (three stool samples on three cards with six windows) while restricting the use of high-dose aspirin or other nonsteroidal antiinflammatory drugs, developing slides within 1 week, and performing a complete colonoscopic evaluation within 2 to 3 months in patients with positive results. We believe that samples should not be rehydrated, but this question has not been definitively resolved. The issue of screening frequency (annual or biennial) involves tradeoffs of benefit and effort that should be analyzed by a quantitative cost-effectiveness analysis. For any method of fecal occult blood testing, the future may bring important improvements, for example, an improved immunochemical assay or a better two-step strategy using a very sensitive test followed by a specific one (for example, Hemoccult-SENSA followed by HemeSelect).

With regard to evaluation after testing, any person older than 50 years of age who has positive results on a fecal occult blood test should generally have a full colonic examination unless it is clear that the source of bleeding is non-neoplastic. It is important for physicians and patients to realize that, although most work-ups will have negative results (that is, colorectal cancer or a significant polyp will not be found), the risk for curable colorectal cancer or a large polyp is high enough after positive results on fecal occult blood testing to warrant a work-up.

Screening with fecal occult blood tests is a surprisingly complex procedure. Thus, although occult blood testing reduces mortality rates associated with colorectal cancer, the effort involved is great, especially with regard to false-positive results and downstream cost and effort. Particular attention is therefore warranted for the details of how screening is done. Clinicians who perform screening with fecal occult blood tests must remain abreast of advances in this field. The key clinical challenge is how to improve the efficiency of the tests by reducing the number of false-negative and especially false-positive results.

3.3 Future Research

The agenda for research into screening with fecal occult blood tests includes several important issues. Screening must be made practical, and compliance must be increased. Physicians' practices and managed care organizations must address systemic and organizational problems to facilitate the implementation of screening programs.

Because it is unlikely that further major clinical trials of screening with fecal occult blood tests will be done with death as the outcome, we expect that important sources of data in the future will include simulated models of the relative diagnostic accuracy of various types of tests [27, 28, 80] or of tests based on the detection of genetic tumor markers instead of bleeding [81]. A long list of research questions has been assembled for screening with fecal occult blood tests (Young G. Meeting of International Union Against Cancer; 1995; Genoa, Italy); in this report, we discuss several that may have the greatest effect on the interpretation of screening results and subsequent patient management.

3.3.1 Improve Compliance

It should be a priority to increase compliance when a screening program using fecal occult blood tests is implemented. Reduction in mortality rates cannot occur unless persons are actually screened. Although the interpretation of post hoc analyses can be problematic, it would be useful to assess the reduction in mortality rates among compliant patients in the three published clinical trials because that reduction probably represents the upper limit of reduction that can be obtained by screening with fecal occult blood tests.

3.3.2 Improve Sensitivity

The success of screening with fecal occult blood tests ultimately depends on the test's ability to detect neoplasms. Therefore, understanding the determinants of sensitivity for detecting neoplasms would be useful for refining screening strategies. To the extent to which important neoplasms (that is, early cancer and large adenomas) do not bleed enough to be detected, the effort to detect occult blood may reflect a preoccupation with an inappropriate target. The most important characteristic of a screening test is high sensitivity [82], but substantial uncertainty still exists about what the sensitivity of occult blood testing is.

Specific questions about sensitivity include the following: To what extent do neoplasms bleed (how frequently and how much) and what features, such as size and histopathology, correlate with bleeding? What is the relation between a test's sensitivity when it is applied at one point in time compared with its sensitivity when it is repeatedly applied in an extended screening program? Does this relation explain the current controversy about whether the sensitivity of Hemoccult for early colorectal cancer is high (about 80% [2]) or low (about 30% [53])? New concepts of sensitivity will be needed to understand and compare not only fecal occult blood tests but also entirely new screening tests [80, 81]. The study of the sensitivity of occult blood tests is methodologically challenging, as discussed above.

3.3.3 Improve Specificity

If the specificity of fecal occult blood tests could be improved without reducing their sensitivity, the cost of unnecessary work-ups could be substantially reduced and the efficiency of the tests could be increased. Improved specificity can theoretically be achieved by using a highly specific test for human hemoglobin that would ignore bleeding in the upper gastrointestinal tract or the presence of peroxidase from vegetable sources. Immunochemical tests provide one important approach to this problem [27, 33]. However, non-neoplastic lower gastrointestinal bleeding (for example, from angiodysplasia) provides a continual source of false-positive results that could be insurmountable.