Performance

FOBT is an indirect test for the presence of cancer or polyps. People with a positive test must undergo some form of diagnostic evaluation to confirm the presence of polyps and/or cancer. The performance and effectiveness of FOBT must therefore be considered in the context of necessary confirmatory diagnostic evaluation.

The positive reaction for blood obtained with the guaiac-based tests results from the pseudoperoxidase activity of hemoglobin. This test is therefore not specific for cancer, since nonneoplastic lesions that cause blood to enter the gastrointestinal tract may result in a positive FOBT. In addition, the test is not specific for blood, since substances with peroxidase or pseudoperoxidase activity (e.g., turnips, horseradish) can cause false-positive reactions if present in the stool.

False-negative tests may occur if cancer or polyps do not bleed while the sample stool is being formed or because the FOBT is not sensitive enough to detect the blood. Vitamin C, an antioxidant, may interfere with the reaction and cause a false-negative test.

The reported sensitivity and specificity of FOBT for colorectal cancer screening vary considerably from study to study. In a study of patients with cancer, sensitivity of nonrehydrated FOBT was 88.8 percent (St. John, Young, Alexeyeff, et al., 1993). In a study of patients under surveillance following curative resection of a colorectal neoplasm, sensitivity was 29 percent for Dukes A and B cancers (Ahlquist, Wieand, Moertel, et al., 1993). This observed wide variation may be due to such factors as whether or not slides were rehydrated, different types of FOBT tests used, the number of samples taken from each patient, whether dietary restrictions were employed, differences in the tests used to confirm diagnosis, whether FOBT was used for prevalence or incidence screening, and differences in patient populations tested. Using FOBT in a program of serial screening over many years will result in improved screening test sensitivity compared with one-time screening, since a program of testing offers many opportunities to detect polyps and cancers that bleed intermittently.

Several prospective controlled trials of screening using the Hemoccult brand of FOBT have been reported (Mandel, Bond, Church, et al., 1993; Kronborg, Fenger, Wonn, et al., 1992; Hardcastle, Armitage, Chamberlain, et al., 1986; Kewenter, Brevinge, Engaras, et al., 1994; Winawer, Flehinger, Schottenfeld, and Miller, 1993; Kronborg, Fenger, Olsen, et al., 1996; Hardcastle, Chamberlain, Robinson, et al., 1996). The trials reflect the performance of a program of repeated screening rather than a single test. In its nonrehydrated form, Hemoccult sensitivity for detecting cancer ranges from 72 to 78 percent, and its specificity is 98 percent. Rehydration - adding a few drops of water before adding the hydrogen peroxide reagent - increases the sensitivity to 88 to 92 percent, with a decrease in specificity from 97 percent to 90 to 92 percent (Mandel, Bond, Church, et al., 1993). Rehydration also results in a decrease in positive predictive value for cancer from 10 to 17 percent to 2 to 6 percent. Thus, for every case of cancer detected, 6 to 10 patients need to undergo a colonoscopy or barium enema as a result of the nonrehydrated test compared with 17 to 50 patients with the rehydrated test.

In a trial of the combination of two newer tests, the Hemoccult II Sensa and HemeSelect, in sequence, sensitivity increased from 38 percent to 65 percent, while specificity remained at 97 percent (Allison, Tekawa, Ransom, and Adrian, 1996). Both of these tests were more sensitive than the nonrehydrated Hemoccult test (St. John, Young, and Alexeyeff, 1993). FOBT is less sensitive for polyps than for cancer, since most polyps do not bleed. However, FOBT screening detects some adenomas because large adenomas are more likely to bleed and because false-positive FOBTs lead to diagnostic testing that discovers polyps whether or not they are a source of bleeding.

Effectiveness

A growing body of evidence has demonstrated that testing for fecal occult blood, with diagnostic evaluation and treatment for positive tests, reduces colorectal cancer mortality. Community trials and case series found that FOBT screening led to cancers being detected at an early pathological stage (Greegor, 1967; Greegor, 1971) and to patients surviving longer after diagnosis (Winawer, Flehinger, Schottenfeld, and Miller 1993).

Three randomized trials of FOBT effectiveness have been completed. Mandel and colleagues (Mandel, Bond, Church, et al., 1993) randomized 46,551 people aged 50 to 80 without symptoms of colorectal cancer into three groups: annual screening, screening every 2 years, and usual care. Initially, nonrehydrated FOBT was performed, but this was replaced by rehydrated FOBT. Trial participants were followed for 13 years. Those with positive FOBTs received colonoscopy and barium enema. The cumulative mortality per 1,000 for colorectal cancer was 5.88 in the annually screened group, 8.33 in the biennially screened group, and 8.83 in the control group, representing a 33 percent reduction in mortality in the group offered annual screening compared with the control group. There was no difference in colorectal cancer incidence among the three groups after 13 years of followup.

Two population-based randomized trials of FOBT have demonstrated significant reductions in colorectal cancer mortality. In the first study by Kronborg and colleagues, 137,485 average-risk people aged 45 to 75 were randomized to either biennial FOBT with the nonrehydrated Hemoccult II or usual health care (Kronborg, Fenger, Olsen, et al., 1996). After 10 years of followup, biennial screening led to an 18 percent reduction in colorectal cancer mortality compared with usual health care. These findings were independent of age and sex. Similarly, in the second study conducted in the United Kingdom, Hardcastle and associates reported a 15 percent reduction in cumulative colorectal cancer mortality using biennial screening with the nonrehydrated FOBT (Hardcastle, Chamberlain, Robinson, et al., 1996).

The findings of these randomized trials are supported by a case-control study (Selby, Friedman, Quesenberry, and Weiss, 1993). The results of these studies reflect the combined effectiveness of screening FOBT with followup diagnostic evaluation (Mandel, Bond, and Church, 1993; Hardcastle, Chamberlain, Robinson, et al., 1996; Kronborg, Fenger, Olsen, et al., 1996). It has been suggested that a portion of the overall benefit of FOBT screening demonstrated in the trial by Mandel and colleagues (1996) is attributable to the diagnostic evaluations received by over one-third of the patients, many because of a high rate of false-positive FOBT results (Lang and Ransohoff, 1994). Evidence from this trial supporting an independent effect of FOBT includes the fact that a significant percentage of patients with an initially positive FOBT were subsequently found to have cancer and the increased likelihood of finding cancer as the number of positive test slides obtained from an individual increased (Mandel, Church, and Ederer, 1993). In the trial by Hardcastle and colleagues, only 4 percent of all individuals who completed FOBT underwent colonoscopy (Hardcastle, Chamberlain, Robinson, et al., 1996); in the trial by Kronborg and associates, 4.3 percent of study participants who were screened at least once underwent colonoscopy (Kronborg, Fenger, Olsen, et al., 1996). As previously stated, statistically significant colorectal cancer mortality reduction was reported in both of these trials.

Screening Frequency

Available evidence suggests that annual screening leads to the greatest reduction in colorectal cancer mortality (Mandel, Bond, Church, et al., 1993). As noted above, evidence from two population-based, randomized controlled trials of biennial FOBT screening also demonstrated a reduction in colorectal cancer mortality (Kronborg, Fenger, Olsen, et al., 1996; Hardcastle, Chamberlain, Robinson, et al., 1996).

Complications

Complications of FOBT result primarily from the diagnostic evaluation of individuals with positive tests. These complications are described in subsequent sections of this report.

Patient Compliance and Participation

Patient compliance rates in studies of FOBT screening vary widely, ranging from 30 percent to 90 percent. In most of these studies, the rate tends to fall off as the study progresses. For example, compliance in one trial was 70 to 80 percent at the beginning, down to 20 percent at 1 year, and 16 percent at 2 years (Winawer, Flehinger, Schottenfeld, and Miller, 1993).

In summary, there is direct evidence that FOBT screening reduces colorectal cancer mortality. However, FOBT screening has limitations. FOBT screening is primarily aimed at detecting cancer, since few adenomatous polyps bleed. The fact that cancers bleed intermittently limits the potential success of FOBT screening for early detection of colorectal cancer. Finally, the use of rehydrated FOBT slides increases sensitivity, but it also increases the false-positive rate and decreases the positive predictive value for colorectal cancer.

Performance

Key questions in assessing the performance of sigmoidoscopy include:

1. How well does it perform in the area of the bowel examined?
2. What percentage of all cancers of the colon and rectum can be evaluated by the instrument?
3. What is the performance of the overall screening strategy that starts with sigmoidoscopy and evaluates abnormal examinations with investigation of the entire bowel?

Available evidence suggests that nearly all cancers and polyps > 1 cm in diameter and 70 to 80 percent of small polyps are identified (Hixson, Fennerty, Sampliner, et al., 1990). False-positive findings are rare, but many biopsied or removed polyps are not adenomatous.

The 60 cm flexible sigmoidoscope reaches up to or beyond the proximal end of the sigmoid colon in 80 percent of examinations and should therefore identify 40 to 60 percent of adenomatous polyps and colorectal cancers (Selby and Friedman, 1989; Selby, Friedman, and Collen, 1988).

If an adenomatous polyp is found on sigmoidoscopy, there is an increased probability that other polyps are present elsewhere in the bowel (Grossman, Milos, Tekawa, and Jewell, 1989; Read, Read, and Butterfly, 1997; Tripp, Morgan, Sampliner, et al., 1987; Winawer, Zauber, O'Brien, et al., 1992). It is common practice to follow an abnormal screening sigmoidoscopy with a full colonoscopy. This practice allows excision of the lesion already found and provides an opportunity to evaluate the colon for more proximal lesions. When adenomatous polyps are found in the rectosigmoid, there is about a one-in-three chance of finding additional adenomas more proximal in the colon (Grossman, Milos, Tekawa, and Jewell, 1989; Read, Read, and Butterfly, 1997; Tripp, Morgan, Sampliner, et al., 1987; Winawer, Zauber, O'Brien, et al., 1992). Further, the characteristics of the proximal adenoma appear to correlate with the characteristics of the rectosigmoid adenomas. When adenomas < 1 cm in diameter are found in the rectosigmoid, it is unlikely that adenomas with advanced pathology will be found proximally (Zarchy and Ershoff, 1994).

Several studies have compared the performance of the 60 cm sigmoidoscope with that of full colonoscopy in the same group of asymptomatic subjects (Foutch, Mai, Pardy, et al., 1991; Lieberman and Smith, 1991; Rex, Lehman, Hawes, et al., 1991). Studies that compared the number of polyps found on the first 60 cm of colonoscope with the number found on full colonoscopy suggest that the 60 cm sigmoidoscope would correctly identify 40 to 60 percent of all adenomas detected by colonoscopy. About one-third of patients with proximal adenomatous polyps had no distal polyps (Lieberman and Smith, 1991; Rex, Lehman, Hawes, et al., 1991). If these patients had undergone only flexible sigmoidoscopy, they would not have been identified as being at increased risk for colorectal cancer.

Effectiveness

There have been no prospective randomized controlled trials addressing the effectiveness or efficacy of screening sigmoidoscopy. The best available evidence of effectiveness in reducing deaths from colorectal cancer comes from three case-control studies (Selby, Friedman, Quesenberry, and Weiss, 1992; Newcomb, Norfleet, Storer, et al., 1992; Muller and Sonnenberg, 1995a). Selby and colleagues (1992) compared the screening histories of people who died of colorectal cancer (cases) with age- and sex-matched controls who were at average risk for colorectal cancer. Only screening with rigid sigmoidoscopy was included; information on performance of sigmoidoscopy was obtained from medical records. The authors adjusted the results for differences in compliance with preventive care in general, taking as their marker the number of periodic health examinations participants in the study had received. They also adjusted for the effect of FOBT. Rigid sigmoidoscopy was associated with a 59 percent reduction in risk of death from cancer in the part of the colon reached by the rigid sigmoidoscope. To test the internal validity of their findings, the authors analyzed those cases and controls who died of cancers that were beyond the reach of the rigid sigmoidoscope. They found that the protective effect of having had a screening sigmoidoscopy was not present for these proximal cancers.

A case-control study by Newcomb and colleagues reported an 80 percent reduction in the risk of death from rectosigmoid cancer in patients who had undergone one or more sigmoidoscopic examinations compared with those who had never done so (Newcomb, Norfleet, Storer, et al., 1992). A study by Muller and Sonnenberg (1995b) included 4,411 U.S. veterans who died of colorectal cancer and both living and dead controls who did not have colorectal cancer. Proctosigmoidoscopy was associated with an odds ratio of 0.41 and 0.40 (living and dead controls, respectively), which represents approximately a 60 percent reduction in risk of death from colorectal cancer in the region viewed.

A number of biases are inherent in the case-control study design. In addition, these case-control studies primarily used rigid rather than flexible sigmoidoscopy and thus provide only indirect evidence of effectiveness of flexible sigmoidoscopy in reducing mortality from colorectal cancer.

Screening Frequency

There are no data from prospective randomized controlled studies to address the question of appropriate frequency for screening sigmoidoscopy. The case-control study of screening sigmoidoscopy by Selby and colleagues found that the protective effect of screening sigmoidoscopy was just as great for those patients who had the procedure 9 to 10 years earlier as for those who had it more recently (Selby, Friedman, Quesenberry, and Weiss, 1992). However, this estimate was based on a retrospective study of data on a small number of patients. Results of the case-control study by Muller and Sonnenberg (1995a) are consistent with a protective effect of at least 6 years.

Complications

The major complication of sigmoidoscopy is colon perforation. Data from large series of sigmoidoscopy show perforation rates ranging from 1 to 2 per 10,000 examinations (Bolt, 1971; Nelson, 1982; Portes and Majarakis, 1957; Selby and Friedman, 1989). Slightly higher complication rates occur when biopsy or polypectomy is performed. No direct evidence of mortality rates for flexible sigmoidoscopy was found in the literature.

Patient Compliance and Participation

Rates of patient participation in screening sigmoidoscopy vary widely. Rates as high as 100 percent have been reported in people more than 50 years of age and with a family history of colorectal cancer (Stephenson, Murday, Finan, et al., 1993). Rates of compliance in studies performed in the workplace range from 31 to 53 percent (Vernon, 1995). The wide variation is due to differences in study design, study population, recruitment methods, and type of sigmoidoscope used. Rates of participation are higher if the procedure is recommended by a physician (Holt, 1991). In a study of who performed screening sigmoidoscopy, people who had the procedure performed by a nurse practitioner were much more likely to appear for retesting than those screened by a physician (Maude, 1994). However, methodologic limitations - including nonrandomized designs and nonrepresentative patient populations - make it difficult to generalize the results of these studies to the population as a whole.

In summary, studies of flexible sigmoidoscopy have demonstrated a reduction in colorectal cancer mortality risk. A disadvantage of sigmoidoscopy screening is that it detects only about half of all colorectal cancers and polyps. There is uncertainty concerning the clinical importance of small adenomatous polyps discovered on sigmoidoscopy screening.

Performance

Available data suggest that the sensitivity of DCBE is 50 to 80 percent for polyps < 1 cm in diameter, 70 to 90 percent for polyps > 1 cm, and 55 to 85 percent for Dukes Stage A and B cancers (Fork, 1981; Steine, Stordahl, Lunde, et al., 1993; Hixson, Fennerty, Sampliner, et al., 1991). Lack of sensitivity is primarily due to inadequate visualization of parts of the bowel and to errors in interpretation. False-positive findings are mainly due to adherent stool and nonneoplastic mucosal irregularities, with rates ranging from less than 1 percent for cancers, to 5 to 10 percent for large polyps, (Steine, Stordahl, Lunde, et al., 1993; Jaramillo and Slezak, 1992; Jensen, Kewenter, Aszteély, et al., 1990), and about 50 percent for small polyps (Steine, Stordahl, Lunde, et al., 1993).

In completed examinations, barium enema images the entire colon. Studies suggest that 5 to 10 percent of examinations are unsatisfactory, requiring another attempt or colonoscopy to visualize the entire colon (Bloomfield, 1981; Jaramillo and Slezak, 1992; Brewster, Grieve, and Saunders, 1994). The performance characteristics of DCBE as a screening test are not known. No prospective randomized trials of DCBE screening have been completed, and most published studies of DCBE relate to symptomatic rather than asymptomatic patients.

DCBE alone has been considered an inaccurate examination for the rectum and sigmoid colon, since adenomas and carcinomas in this area are sometimes overlooked. Because flexible sigmoidoscopy can examine the rectum and sigmoid colon, the combination of these two tests can lead to improved detection of adenomatous polyps and cancer. A randomized trial of DCBE plus flexible sigmoidoscopy versus colonoscopy in 383 patients with gastrointestinal bleeding suspected to be from the colon found that colonoscopy detected more cases of polyps < 9 mm than DCBE plus sigmoidoscopy (Rex, Weddle, and Lehman, 1990). There was no difference between strategies in the number of patients detected with cancer or polyps > 9 mm. This was a study of symptomatic patients and is not generalizable to asymptomatic populations.

DCBE combined with flexible sigmoidoscopy is the diagnostic followup evaluation in a large ongoing trial of screening FOBT being conducted in Sweden (Kewenter, Brevinge, Engaras, and Haglind, 1995). In 1,831 patients who underwent full examination, 25 percent of carcinomas and a similar proportion of adenomas > 1 cm in diameter in the rectosigmoid area were overlooked by DCBE. Flexible sigmoidoscopy overlooked 8 percent of carcinomas and 10 percent of adenomas in the rectosigmoid. Overlooked carcinomas were defined as those diagnosed within 1 year of complete examination. In this study, 21 carcinomas beyond the reach of the flexible sigmoidoscope were diagnosed with DCBE. The combination of DCBE and flexible sigmoidoscopy had a sensitivity of 98 percent for carcinomas and 99 percent for adenomas. These data suggest that adding flexible sigmoidoscopy to DCBE provides complementary information.

Effectiveness and Screening Frequency

Literature review revealed no controlled studies of the effectiveness of DCBE for colorectal cancer screening in which death or any other adverse health outcome from colorectal cancer was measured. Similarly, there are no studies that directly address the question of how frequently DCBE should be performed to screen for colorectal cancer and adenomatous polyps. Indirect evidence of DCBE's effectiveness in screening is that detecting polyps and early cancers reduces colorectal cancer mortality, and DCBE detects many of these lesions.

Complications

The most serious complication of DCBE is bowel perforation. Data on how often perforations occur, as well as the frequency of other potential complications, are sparse.

Patient Compliance and Participation

Studies on acceptability of DCBE among patients being investigated for suspected colonic disease give inconsistent results. One found that 94 percent of patients rated DCBE as an acceptable procedure, while nearly half of the patients in another study found DCBE to be distressing and uncomfortable (Stein, 1994; Williams, Macrae, and Bartram, 1982; Durdey, Weston, and Williams, 1987). Few data are available on patient participation in use of DCBE in screening settings.

In summary, there is no direct evidence that DCBE can reduce mortality from colorectal cancer. Screening DCBE can image the entire colon and detect cancers and large polyps almost as well as colonoscopy. It is imprecise, however, in identifying small polyps. The combination of flexible sigmoidoscopy with DCBE provides complementary information and increases sensitivity for detection of cancer and polyps.

Performance

As with many procedures, colonoscopy techniques have improved since the procedure was first introduced in the early 1970s. However, procedural competence varies (Baille and Ravich, 1993). The cecum is reached in 80 to 98 percent of procedures (Rex, Lehman, Hawes, et al., 1991; Lieberman and Smith, 1991; Godreau, 1992), the depth of penetration depending mainly on the experience of the endoscopist and the adequacy of bowel preparation (Anderson, Heigh, McCoy, et al., 1992; Cass, Freeman, Peine, et al., 1993). Most available data on performance are from diagnostic evaluations and surveillance rather than from screening. In one study of screening colonoscopy, 98.6 percent of examinations reached the cecum (Rex, Lehman, Hawes, et al., 1991).

Although colonoscopy can detect cancers and polyps, it is less accurate in detecting small polyps. A limitation of many studies evaluating the performance of colonoscopy is that colonoscopy itself is taken as the gold standard for the presence or absence of polyps and cancers. A retrospective review of 429 patients who had colorectal cancer and polyps resected after undergoing preoperative colonoscopy showed that the findings at colonoscopy correlated with the pathological specimen in 97 percent of cases, but that colonoscopy missed the lesion in the remaining 3 percent of cases (Byrd, Boggs, Slagle, and Cole, 1989). A retrospective study conducted in 1992 by Warneke and colleagues included patients in a cancer referral center who had undergone preoperative colonoscopy and subsequently underwent colectomy for primary colorectal cancer between 1980 and 1987. In these patients, 36 of 46 polyps > 1 cm located in the area examined during colonoscopy were found by colonoscopy (Warneke, Petrelli, Herrera, and Nava, 1992). A prospective study, in which 90 patients underwent colonoscopy by two experienced examiners, concluded that large polyps (> 1 cm) were rarely missed, but the operator missed about 15 percent of small polyps (Hixson, Fennerty, Sampliner, et al., 1990). Other studies have used followup colonoscopy in a short time frame - 6 months or 1 year - as the gold standard for the existence of polyps (Hoff, Foertser, Vatn, et al., 1986; Waye, Lewis, Frankel, and Geller, 1988). These findings suggest that colonoscopy misses 25 percent of polyps < 5 mm and 10 percent of polyps > 1 cm.

False-positive results are rare, although about one-third of polyps removed are not adenomatous (O'Brien, Winawer, Zauber, et al., 1990).

Effectiveness

There are no studies that examine the effectiveness of colonoscopy as a screening test for colorectal cancer with reduction in colorectal cancer mortality as a study endpoint. Winawer and colleagues demonstrated that detecting and removing polyps reduced the incidence of colorectal cancer (Winawer, Zauber, Ho, et al., 1993). Trials of FOBT effectiveness demonstrate that detecting early cancers and precancerous lesions lowers colorectal cancer mortality. Colonoscopy detects most of these lesions. It has been suggested that to the extent that colonoscopy is a part of the intervention in FOBT screening trials, evidence of screening intervention effectiveness is indirect evidence of colonoscopy effectiveness (Mandel, Bond, Church, et al., 1993).

Screening Frequency

There are no controlled trials that address the question of how frequent colonoscopy should be performed to screen for colorectal cancer.

Complications

Complications of colonoscopy include perforation, hemorrhage, respiratory depression due to sedation, arrhythmia during the procedure, transient abdominal pain and ileus, and nosocomial infection. Data from six prospective studies of colonoscopy indicate that approximately 1 in 1,000 patients suffers perforation, 3 in 1,000 suffer major hemorrhage, and 1 to 3 in 10,000 die as a result of the procedure (Jorgensen, Kronborg, and Fenger, 1993; Waye, Lewis, and Yessayan, 1992; Jentschura, Raute, Winter, et al., 1994; Rex, Lehman, Hawes, et al., 1991; McAfee and Katon, 1994; Godreau, 1992). Complication rates may be higher if polypectomy is performed (Rosen, Bub, Reed, and Nastasee, 1993). Also, about 5 in 1,000 patients experience clinically significant respiratory depression (Kalra and Hamlyn, 1988).

Patient Compliance and Participation

Data on compliance with screening colonoscopy are sparse. When physicians, nurses, and spouses of the nurses and physicians were invited by letter to undergo free screening colonoscopy, less than 15 percent accepted (Rex, Lehman, Ulbright, et al., 1993). In the National Polyp Study, 80 percent of people appeared for repeat testing after previous polypectomy (Winawer, Zauber, O'Brien, et al., 1992). Opinions about acceptability vary. Williams and colleagues found that 88 percent of patients rated colonoscopy as an acceptable procedure (Williams, Macrae, and Bartram, 1982). Nearly one-fourth of patients studied by Durdey and colleagues found the procedure to be distressing and uncomfortable (Durdey, Weston, Williams, 1987).

In summary, there are no controlled trials evaluating the effectiveness of screening colonoscopy in reducing colorectal cancer mortality in people at average risk for the disease. Colonoscopy was an integral part of FOBT screening trials that demonstrated a significant reduction in mortality in screened patients. In addition, colonoscopy has been shown to decrease colorectal cancer incidence in a cohort of patients with adenomatous polyps. Colonoscopy permits visualization of the entire colon in most cases, detection and removal of polyps, and biopsy of lesions suggestive of cancer.