Fecal Occult Blood Test (FOBT)
In FOBT testing, stool samples are collected and analyzed for the presence of small amounts of blood. The first generation of FOBTs used guaiac-based assays to detect blood, which are less sensitive and less specific than immunochemical-based testing. The now-classic randomized controlled trials (RCTs) that assessed colorectal cancer (CRC) mortality reduction all used guaiac-based testing. The finding of decreased CRC mortality provided a major foundation to do CRC screening recommendations. The first-generation guaiac-based tests are being replaced with more sensitive and more specific immunochemical tests, which have not yet been assessed in RCTs with a no-screening control group.
In this setting, the RCT evidence about guaiac-based testing is reviewed briefly here, with further discussion of how immunochemical FOBT (iFOBT or FIT) may provide improved sensitivity and specificity. Generally, if guaiac FOBT (gFOBT) is acceptable as a screening test (as shown in RCTs), then a strong case can be made for using a more sensitive and more specific test like FIT.
gFOBT collection details vary somewhat for different tests, but they typically involve collection of as many as three different specimens on 3 different days, with small amounts from one specimen smeared by a wooden stick on a card with two windows or otherwise placed in a specimen container.
The guaiac test identifies peroxidase-like activity that is characteristic of human and nonhuman hemoglobin. Thus, the test records blood from ingested meat, upper airway bleeding such as epistaxis, upper gastrointestinal (GI) bleeding, and colonic lesions.
A systematic review regarding evidence of benefit was conducted through the Cochrane Collaboration. It examined all CRC screening randomized trials that involved gFOBT testing done on more than one occasion. The combined results showed that trial participants allocated to screening had a 16% lower CRC mortality (relative risk , 0.84; 95% confidence interval , 0.78–0.90). There was no difference in all-cause mortality between the screened groups and the control groups (RR, 1.00; 95% CI, 0.99–1.02). The trials reported a low positive predictive value (PPV) for the FOBT test, suggesting that most positive tests were false positives. The PPV was 5.0% to 18.7% in the trials using nonrehydrated slides (Funen and Nottingham studies), and it was 0.9% to 6.1% in the trials using rehydrated slides (Goteborg and Minnesota studies). The report contained no discussion about contamination in the control arms of the trials and no information about treatment by disease stage.
On initial (prevalence) examinations, 1% to 5% of unselected individuals tested with gFOBT had positive test results. Of those who tested positive, approximately 2% to 10% had cancer and approximately 20% to 30% had adenomas, depending on how the test was done. Data from RCTs of gFOBT testing are summarized in Table 2.
Four controlled clinical trials to evaluate the efficacy of screening with gFOBT have been completed or are in progress. While more sensitive stool blood tests based on measuring human hemoglobin have been developed (and are discussed later in this summary), results of their performance in RCTs have not yet been reported. For gFOBT, the Swedish trial was a targeted study for individuals aged 60 to 64 years. The English trial selected candidates from lists of family practitioners. The Danish trial offered screening to a population aged 45 to 75 years who were randomly assigned to a control or study group.
The Minnesota trial randomly assigned 46,551 men and women aged 50 to 80 years to one of three arms: CRC screening with gFOBT, rehydrated (with some small percentage of unrehydrated) FOBT every year (n = 15,570) or every 2 years (n = 15,587), or a control group (n = 15,394). This trial demonstrated that annual FOBT screening decreased mortality from CRC by 33% after 18 years of follow-up (RR, 0.67; 95% CI, 0.51–0.83, compared with the control group) and that biennial testing resulted in a 21% relative mortality reduction (RR, 0.79; 95% CI, 0.62–0.97). Some part of the reduction may have been attributed to chance detection of cancer by colonoscopies; rehydration of guaiac test slides greatly increased positivity and consequently increased the number of colonoscopies performed. Subsequent analyses by the Minnesota investigators using mathematical modeling suggested that for 75% to 84% of the patients, mortality reduction was achieved because of sensitive detection of CRCs by the test; chance detection played a modest role (16%–25% of the reduction). Nearly 85% of patients with a positive test underwent diagnostic procedures that included colonoscopy or double-contrast barium enema plus flexible sigmoidoscopy (FS). After 18 years of follow-up, the incidence of CRC was reduced by 20% in the annually screened arm and 17% in the biennially screened arm. With follow-up through 30 years, there was a sustained reduction in CRC mortality of 32% in the annually screened arm (RR, 0.68; 95% CI, 0.56–0.82) and 22% in the biennially screened arm (RR, 0.78; 95% CI, 0.65–0.93). There was no reduction in all-cause mortality in either screened arm (RR, 1.00; 95% CI, 0.99–1.01 for the annually screened arm and RR, 0.99; 95% CI, 0.98–1.01 for the biennially screened arm). Important information that was not reported includes the treatment of CRC cases by stage by arm and the extent of CRC screening in each arm by FOBT, sigmoidoscopy, or colonoscopy after the completion of the trial protocol.
The English trial allocated approximately 76,000 individuals to each arm. Those in the screened arm were offered nonrehydrated gFOBT testing every 2 years for three to six rounds from 1985 to 1995. At a median follow-up of 7.8 years, 60% completed at least one test, and 38% completed all tests. Cumulative incidence of CRC was similar in both arms, and the trial reported an RR reduction of 15% in CRC mortality (odds ratio , 0.85; 95% CI, 0.74–0.98). The serious complication rate of colonoscopy was 0.5%. There were five deaths within 30 days of surgery for screen-detected CRC or adenoma in a total of 75,253 individuals screened. After a median follow-up of 11.8 years, no difference in CRC incidence between the intervention and control groups was observed. The disease-specific mortality rate ratio associated with screening was 0.87 (0.78–0.97; P = .01). The rate ratio for death from all causes was 1.00 (0.98–1.02; P = .79). When the median follow-up was extended to 19.5 years, there was a 9% reduction in CRC mortality (RR, 0.91; 95% CI, 0.84–0.98) but no reduction in CRC incidence (RR, 0.97; 95% CI, 0.91–1.03), or death from all causes (RR, 1.00; 95% CI, 0.99–1.02).
The Danish trial in Funen, Denmark, entered approximately 31,000 individuals into two arms, in which individuals in the screened arm were offered nonrehydrated gFOBT testing every 2 years for nine rounds over a 17-year period. Sixty-seven percent of participants completed the first screen, and more than 90% of individuals invited to each subsequent screen underwent FOBT testing. This trial demonstrated a reduction of 18% in CRC mortality at 10 years of follow-up, 15% at 13 years of follow-up (RR, 0.85; 95% CI, 0.73–1.00), and 11% at 17 years of follow-up (RR, 0.89; 95% CI, 0.78–1.01). CRC incidence and overall mortality were virtually identical in both arms.
The Swedish trial in Goteborg enrolled all of its 68,308 citizens in the city who were born between 1918 and 1931 and were aged 60 to 64 years, and randomly assigned them to screening and control groups of nearly equal size. Participants in the control group were not contacted and were unaware they were part of the trial. Screening was offered at different frequencies to three different cohorts according to year of birth. Screening was done using the gFOBT Hemoccult-II test after dietary restriction. Nearly 92% of tests were rehydrated. Individuals with a positive test result were invited to an examination consisting of a case history, FS, and double-contrast barium enema. Follow-up ranged from 6 years 7 months to 19 years 5 months, depending on the date of enrollment. The primary end point was CRC-specific mortality. The overall screening compliance rate was 70%, and 47.2% of participants completed all screenings. Of the 2,180 participants with a positive test, 1,890 (86.7%) underwent a complete diagnostic evaluation with 104 cancers and 305 adenomas of at least 10 mm detected. In total, there were 721 CRCs (152 Dukes D, 184 Dukes C) in the screening group and 754 CRCs (161 Dukes D, 221 Dukes C) in the control group, with an incidence ratio of 0.96 (95% CI, 0.86–1.06). Deaths from CRC were 252 in the screening group and 300 in the control group, with a mortality ratio of 0.84 (95% CI, 0.71–0.99). This CRC mortality difference emerged after 9 years of follow-up. Deaths from all causes were very similar in the two groups, with a mortality ratio of 1.02 (95% CI, 0.99–1.06).
Stage distribution
All trials have shown a more favorable stage distribution in the screened population than in controls (see Table 2). Data from the Danish trial indicated that while the cumulative incidence of CRC was similar in the screened and control groups, a higher percentage of CRCs and adenomas were Dukes A and Dukes B lesions in the screened group. A meta-analysis of all previously reported randomized trials using biennial FOBT showed no overall mortality reduction by gFOBT screening (RR, 1.002; 95% CI, 0.989–1.085). The RR of CRC death in the gFOBT arm was 0.87 (95% CI, 0.8–0.95), and the RR of non–CRC death in the gFOBT group was 1.02 (95% CI, 1.00–1.04; P = .015).
Mathematical modeling
Mathematical models have been constructed to extrapolate the results of screening trials and screening programs for benefit of the general population in community health care delivery settings. These models project that using currently available screening methodology can reduce CRC mortality or increase life expectancy.
Table 2. Randomized Controlled Screening Trials to Assess Outcome: Guaiac-Based Fecal Occult Blood Testing
| Site | Population Size | Positivity Rate (%) | % Cancers Localizeda | Testing Interval | CRC Mortality Relative Risk (95% CI) | CRC Incidence RR (95% CI) |
|---|
| CI = confidence interval; CRC = colorectal cancer; RR = risk ratio. |
| a% Localized = T1–3 N0 M0. |
| | Screened | Control | | |
| Minnesota | 48,000 | Unrehydrated: 2.4% | 59 | 53 | Annual | 0.67 (0.51–0.83) | 0.80 (0.70–0.90) |
| | Rehydrated: 9.8% | | | Biennial | 0.79 (0.62–0.97) | 0.83 (0.73–0.94) |
| United Kingdom | 150,000 | Unrehydrated: 2.1% | 52 | 44 | Biennial | 0.85 (0.74–0.98) | 1.04 (0.95–1.14) |
| Denmark | 62,000 | Unrehydrated: 1.0% | 56 | 48 | Biennial | 0.82 (0.68–0.99) | 1.00 (0.87–1.13) |
| Sweden | 68,308 | Unrehydrated: 1.9% | 52 | 50 | Varied | 0.84 (0.71–0.99) | 0.96 (0.86–1.06) |
| | Rehydrated: 5.8% | | | | | |
Immunochemical FOBTs (iFOBT or FIT): Nonrandomized Controlled Trial Evidence to Assess Lesion Detection
The immunochemical FOBT (iFOBT or FIT) was developed to detect intact human hemoglobin. The advantage of FIT over gFOBT is that it does not detect hemoglobin from nonhuman dietary sources. Also, FIT does not detect partly digested human hemoglobin that comes from the upper respiratory or GI tract. Preliminary studies of several commercially developed FIT tests define their sensitivity and specificity compared with concurrently performed colonoscopy. The studies also examine these outcomes for different cutpoints, and the benefit of multiple versus single stool samples.
Overall, FIT testing is much more sensitive than gFOBT, and it is more sensitive for cancers than for benign neoplasias. As expected, higher cutpoints decrease sensitivity and increase specificity. Fecal immunochemical tests may vary with regard to numbers of stools tested and cutoff values for a positive result.
A systematic review of FIT studies in 2019 found 31 studies, with 120,255 participants and 18 types of FIT tests, that used screening colonoscopy as the reference standard, thus allowing calculation of test sensitivity and specificity. Performance depended on the threshold for a positive result, so that a threshold of 10 µg/g (micrograms of hemoglobin per gram of feces) resulted in a CRC sensitivity of 0.91 (95% CI, 0.84–0.95) and a specificity of 0.90 (95% CI, 0.86–0.93), while a threshold greater than 20 µg/g resulted in a sensitivity of 0.71 (95% CI, 0.56–0.83) with specificity of 0.95 (95% CI, 0.94–0.96). For advanced adenomas, at a threshold of 10 µg/g, sensitivity was 0.40 (95% CI, 0.33–0.47) with a specificity of 0.90 (95% CI, 0.87–0.93). Comparison of three FITs at three thresholds was inconclusive because CIs overlapped, and the comparisons were across rather than within studies. Overall, FIT appears to provide a substantially improved sensitivity compared with gFOBT, although with some compromise in specificity.
The diagnostic sensitivity of FIT testing may vary depending on lesion location in the colon. Proximal lesions may be harder to detect for several reasons, including that they may arise from serrated lesions that are flat and, because they are less vascular than traditional adenomas, tend to bleed less frequently. In a population-based screening program of every-other-year FIT (set to detect 100 ng of hemoglobin per mL of buffer) testing, individuals who had six FITs over time were assessed to learn the frequency with which proximal and distal lesions were discovered. Over 12 years (2002–2014), 123,000 participants had 441,000 FITs. The detection rate for proximal colon cancer declined only from the first to the second screening round (0.63–0.36 per 1,000 screened participants), while the rate for both distal colon and rectal cancer decreased across all six rounds (distal cancer, 1.65 in the first round to 0.17 in the sixth round). (Similar trends occurred for advanced adenomas.) The proportional interval cancer rate—the number of cancers observed versus expected—was higher in the proximal colon than in the distal colon (25.2% vs. 6.0%), suggesting that many proximal cancers (or their immediate precursors) may have been missed by FIT. These results suggest that FIT is less sensitive for proximal CRC and certainly for advanced adenomas, although it is possible that the miss rate may have been inflated if colonoscopy done in response to a positive FIT had missed a precursor lesion. Overall, these results raise questions about the degree of efficacy of FIT in preventing proximal CRC mortality.
The performance and acceptability of FIT over time was assessed by Kaiser-Permanente of Northern and Southern California in a screening program. A retrospective cohort of 323,349 persons aged 50 to 70 years was followed for up to four screening rounds over 4 years. Of patients invited, participation in round one was 48.2%, and of those remaining eligible, 75.3% to 86.1% participated in subsequent rounds. The authors reported that “programmatic FIT screening detected 80.4% of patients with CRC diagnosed within 1 year of testing, including 84.5% in round one and 73.4% to 78.0% in subsequent rounds.” An important observation was the degree of participation found. One limitation of the study is that it was not clear how work-up bias was addressed; e.g., when individuals with a positive test result are preferentially worked up to ascertain the presence or absence of CRC, while individuals with a negative test, but who might have CRC, are not. Although a look-back method was used to ascertain whether an individual had cancer, it is not clear that the duration of follow-up was long enough to discover everyone who should have been included in the denominator of the sensitivity calculation. Nevertheless, the results suggested that subsequent FIT results were at least partially independent of previous results. Longer follow-up may help clarify this issue. Mortality reduction could not be assessed in this study.
Potential false-positive test results because of an increased risk of upper GI bleeding are of concern with FOBT testing and pretest protocols, therefore; low-dose aspirin regimens are discontinued for a week or more before FOBT. The performance of FIT was tested in an ongoing diagnostic study (2005–2009) at 20 internal medicine GI practices in southern Germany. Nineteen hundred seventy-nine patients (233 regular low-dose aspirin users and 1,746 never users) were identified in the records for inclusion in the analysis. All patients provided one stool sample taken within a week before colonoscopy preparation, which was collected according to instructions in a container that was kept refrigerated or frozen until rendered to the clinic on the day of colonoscopy, and the patients agreed to complete a standard questionnaire regarding the use of analgesics and low-dose aspirin (for prevention of cardiovascular disease). Stool samples were thawed within a median of 4 days after arrival at the central laboratory (shipped frozen from the recipient clinics). Fecal occult blood levels were measured by two automated FIT tests according to the manufacturer’s instructions (RIDASCREEN Haemoglobin and RIDASCREEN Haemo-/Haptoglobin Complex, r-biopharm, Bensheim, Germany) following clinical procedures and blinded to colonoscopy results. Advanced neoplasms were found in 24 aspirin users (10.3%) and in 181 nonusers (10.4%). At the cut point recommended by the manufacturer, sensitivities for the two tests were 70.8% (95% CI, 48.9%–87.4%) for users compared with 35.9% (95% CI, 28.9%–43.4%) for nonusers and 58.3% (95% CI, 36.6%–77.9%) for users compared with 32% (95% CI, 25.3%–39.4%) for nonusers (P = .001 and P = .01, respectively). Specificities were 85.7% (95% CI, 80.2–90.1%) for users compared with 89.2% (95% CI, 87.6%–90.7%) for nonusers and 85.7% (95% CI, 80.2%–90.1%) for users compared with 91.1% (95% CI, 89.5%–92.4%) for nonusers (P = .13 and P = .01, respectively). For these FITs, sensitivity for advanced neoplasms was notably higher with the use of low-dose aspirin while specificity was only slightly reduced, suggesting that there might be an advantage of aspirin use to increase sensitivity without much decrease in specificity.
Sigmoidoscopy
The flexible fiberoptic sigmoidoscope was introduced in 1969. The 60 cm flexible sigmoidoscope became available in 1976. The flexible sigmoidoscope permits a more complete examination of the distal colon with more acceptable patient tolerance than the older rigid sigmoidoscope. The rigid instrument can discover 25% of polyps, while the 60-cm scope can find as many as 65% of them. The finding of an adenoma by FS may warrant a colonoscopy to evaluate the more proximal portion of the colon. The prevalence of advanced proximal neoplasia is increased in patients with a villous or tubulovillous adenoma distally and is also increased in those aged 65 years or older with a positive family history of CRC and with multiple distal adenomas. Most of these adenomas are polypoid, flat, and depressed lesions, which may be somewhat more prevalent than previously recognized.
Four major sigmoidoscopy screening RCTs have reported incidence and mortality results (a fifth, the Telemark trial in Norway, was very small, with 800 total participants). These are the Norwegian Colorectal Cancer Prevention (NORCCAP) trial; the United Kingdom Flexible Sigmoidoscopy Screening Trial (UKFSST); the Screening for COlon REctum (SCORE) trial in Italy; and the U.S. Prostate, Lung, Colorectal, and Ovarian (PLCO) Cancer Screening Trial (see Table 3). Participants were aged 55 to 74 years in PLCO and aged 55 to 64 years in the other three trials. Together, the trials enrolled 166,000 participants in the screened groups and 250,000 participants in the control groups. Median follow-up was approximately 11 years for each group. Results were summarized in three systematic reviews. There was an 18% relative reduction in CRC incidence (RR, 0.82; 95% CI, 0.75–0.89), an overall 28% relative reduction in CRC mortality (RR, 0.72; 95% CI, 0.65–0.80), a 31% relative reduction in the incidence of distal CRC (RR, 0.69; 95% CI, 0.63–0.74), and a 46% relative reduction in the mortality of distal CRC (RR, 0.54; 95% CI, 0.43–0.67). A meta-analysis showed a statistically significant, although clinically small, effect on all-cause mortality (RR, 0.97; 95% CI, 0.96–0.99).
Three of the four trials above published long-term follow-up analyses of trial results. For UKFSST, median follow-up was 17.1 years. The RRs for CRC incidence and mortality were similar to those originally reported: an RR of 0.70 (95% CI, 0.62–0.79) for CRC mortality and an RR of 0.74 (95% CI, 0.70–0.80) for CRC incidence. For the PLCO trial, median follow-up was 15.8 years for incidence and 16.8 years for mortality; RRs were 0.75 (95% CI, 0.66–0.85) for CRC mortality and 0.82 (95% CI, 0.76–0.88) for CRC incidence. Median follow-up in the NORCCAP trial was approximately 15 years; HRs were 0.79 (95% CI, 0.65–0.96) for CRC mortality and 0.78 (95% CI, 0.70–0.87) for CRC incidence. A pooled analysis of the four trials, using data through 15 years of follow-up and restricted to those aged 55 to 64 years at randomization, showed a rate ratio of 0.79 (95% CI, 0.75–0.83) for CRC incidence and 0.80 (95% CI, 0.72–0.88) for CRC mortality.
There are no strong direct data from studies of sigmoidoscopy to determine the optimal frequency of screening tests in programs of screening.
Table 3. Randomized Controlled Screening Trials to Assess Outcome: Sigmoidoscopya
| Site | Population Size (Intervention) | FS Rate (%)b | Colonoscopy Rate (%)c | Cumulative CRC Incidence (%) | CRC Deaths per 100,000 Person-Years | CRC Mortality Relative Risk (95% CI) | Distal CRC Mortality Relative Risk (95% CI)e | CRC Incidence Relative Risk (95% CI) |
|---|
| CI = confidence interval; CRC = colorectal cancer; FS = flexible sigmoidoscopy. |
| aAdapted from Lin et al. |
| bThe FS rate refers to the % of individuals who received FSG in the screened group. |
| cThe colonoscopy rate refers to the % of individuals who received a colonoscopy as a follow-up to a positive sigmoidoscopy among those who received a sigmoidoscopy. In the U.S. study, individuals with a polyp found at the time of a sigmoidoscopy were referred for diagnostic follow-up, which was generally done with a colonoscopy. In the other studies, the referral criteria for a colonoscopy depended on the histology of lesion(s) found at the time of the sigmoidoscopy. |
| d Half of the intervention group was also offered FOBT. |
| eThese data are from Lin et al., eFigure 1. Forest plot of randomized controlled trials of FS screening on distal CRC mortality. |
| United Kingdom 2010 | Intervention: 57,099 | 71.1 | 5.0 | 1.5 | Intervention: 30 | 0.69 (0.59–0.80) | 0.58 (0.46–0.74) | 0.77 (0.70–0.84) |
| Control: 112,939 | Control: 44 |
| Italy 2011 | Intervention: 17,136 | 57.8 | 7.8 | 1.6 | Intervention: 35 | 0.78 (0.56–1.08) | 0.73 (0.47–1.12) | 0.82 (0.69–0.96) |
| Control: 17,136 | Control: 44 |
| United States 2012 | Intervention: 77,445 | 86.6 | 25.3 | 1.5 | Intervention: 29 | 0.74 (0.63–0.87) | 0.50 (0.38–0.64) | 0.79 (0.72–0.85) |
| Control: 77,455 | Control: 39 |
| Norway 2014 | Intervention: 20,572d | 63.0 | 19.5 | 1.4 | Intervention: 31 | 0.73 (0.56–0.94) | 0.87 (0.61–0.043) | 0.80 (0.70–0.92) |
| Control: 78,220 | Control: 43 |
Combination of FOBT and Flexible Sigmoidoscopy: Impact on Neoplasm Detection
A combination of FOBT and sigmoidoscopy might increase the detection of lesions in the left colon (compared with sigmoidoscopy alone) while also increasing the detection of lesions in the right colon. Sigmoidoscopy detects lesions in the left colon directly but detects lesions in the right colon only indirectly when a positive sigmoidoscopy (that may variously be defined as a finding of advanced adenoma, any adenoma, or any polyp) is used to trigger a colonoscopic examination of the whole colon.
In 2,885 veterans (97% male; mean age, 63 years), the prevalence of advanced adenoma at colonoscopy was 10.6%. It was estimated that combined screening with one-time FOBT and sigmoidoscopy would detect 75.8% (95% CI, 71.0%–80.6%) of advanced neoplasms. Examination of the rectum and sigmoid colon during colonoscopy was defined as a surrogate for sigmoidoscopy. This represented a small but statistically insignificant increase in the rate of detection of advanced neoplasia when compared with FS alone (70.3%; 95% CI, 65.2%–75.4%). The latter result could be achieved assuming that all patients with an adenoma in the distal colon undergo complete colonoscopy. Advanced neoplasia was defined as a lesion measuring at least 10 mm in diameter, containing 25% or more villous histology, high-grade dysplasia, or invasive cancer. One-time use of FOBT differs from the annual or biennial application reported in those studies summarized in Table 2.
A study of 21,794 asymptomatic individuals (72% were men) who had both colonoscopy and FIT for occult blood compared the detection of proximal cancers as triggered by different test results. FIT alone resulted in a sensitivity of 58.3% and a specificity of 94.5% for proximal cancer diagnosis. FIT plus the finding of advanced neoplasia in the rectosigmoid colon yielded a sensitivity of 62.5% and a specificity of 93%. In this study, the addition of sigmoidoscopy to FIT did not substantially improve the detection of proximal colon cancers, compared with FIT alone.
Colonoscopy
Randomized controlled trial evidence about colorectal cancer incidence or mortality reduction
Five RCTs exploring the impact of colonoscopy on CRC mortality or incidence have been initiated (NCT01239082, NCT00883792, NCT00906997, NCT02078804, and ChiCTR1900025257). Results of one RCT have been reported.
The NordICC trial (NCT00883792) was a pragmatic randomized trial of colonoscopy screening, begun in 2009 in Poland, Norway, Sweden, and the Netherlands. Men and women aged 55 to 64 years were identified from population registries and randomly assigned in a 1:2 ratio to receive either an invitation to undergo a single colonoscopy screening (invited group) or no invitation to screening (usual care group). Individuals were randomized before they were asked whether they wanted to participate in the trial. Except for the 6,900 participants from Norway, participants in the usual-care group were not informed of their enrollment in the trial. The primary endpoints were risk of developing CRC and CRC-related death. The secondary endpoint was death from any cause.
Data were reported for 84,585 participants in Poland, Norway, and Sweden: 28,220 in the invited group, of whom 42% underwent screening, and 56,365 in the usual-care group. Opportunistic colonoscopy screening activity was monitored in the trial regions, and no additional colonoscopy procedures were identified beyond what would have been expected for clinical indications, indicating minimal, if any, contamination in the usual-care group. Hence, this trial compared colonoscopy screening with no screening. After a median follow-up of 10 years, in intention-to-screen analyses, the risk of developing CRC was 0.98% in the invited group and 1.20% in the usual-care group (risk ratio, 0.82; 95% CI, 0.70–0.93). The risk of death from CRC was 0.28% in the invited group and 0.31% in the usual-care group (risk ratio, 0.90; 95% CI, 0.64–1.16). The risk of death from any cause was 11.03% in the invited group and 11.04% in the usual-care group (risk ratio, 0.99; 95% CI, 0.96–1.04). Since only 42% of invited participants underwent screening, adjusted analyses were performed to estimate the effect of screening if all participants randomly assigned to screening actually had been screened. The risk ratio for CRC was 0.69 (95% CI, 0.55–0.83), and the risk ratio for CRC-related death was 0.50 (95% CI, 0.27–0.77). Fifteen participants had major bleeding after polyp removal, and there were no perforations or screening-related deaths within 30 days after colonoscopy.
Caveats for this trial include the following:
- Data from the Netherlands, roughly 10% of the participants, were not included.
- Only 42% of invited participants underwent screening.
- Follow-up duration may not be long enough to realize the full effect of the screening; additional analyses after 15 years are anticipated.
- Colonoscopy is operator dependent, and 29% of the trial endoscopists had an adenoma detection rate (ADR) below the recommended minimum threshold.
- There is a suggestion that high-risk individuals in Poland chose to undergo colonoscopy, which could have led to an underestimation of the screening effect.
Indirect evidence of benefit comes from the detection rate of lesions that may be clinically important (like early CRC or advanced adenomas). Some case-control results are available. The sensitivity of a CRC screening test for adenomas (and for CRC) may be helpful in considering its possible clinical usefulness, given that there are no completed RCTs of the impact of colonoscopy on CRC mortality or incidence. Colonoscopy is commonly considered the gold standard because it directly assesses the physical presence of lesions in the colon. However, colonoscopy can miss roughly 10% of cancers and advanced adenomas because of suboptimal bowel cleansing, lesions being hidden behind folds (or haustra) in the colon, or suboptimal examination by the endoscopist. Recent data suggest that the magnitude of a endoscopist's adenoma-detection rate (commonly measured as the proportion of colonoscopies in which an adenoma is found) is related to reduced incidence of CRC.
Adenoma detection rate (ADR)
Detection rates in colonoscopy screening vary with the rate at which the endoscopist examines the colon while withdrawing the scope. In one study, there were differences among gastroenterologists in the rates of detection of adenomas (range of the mean number of lesions per patient screened, 0.10–1.05; range of the percentage of patients with adenomas, 9.4%–23.7%) and the times of withdrawal of the scope (3.1–16.8 minutes for procedures not including polyp removal). Examiners whose mean withdrawal time was 6 minutes or more had higher detection rates than those with mean withdrawal times of less than 6 minutes (28.3% vs. 11.8%; P < .001 for any neoplasia and 6.4% vs. 2.6%; P < .005 for advanced neoplasia).
In the first 10 years of the German CRC screening program, detection of nonadvanced adenomas increased in men from 13.3% to 22.3% and in women from 8.4% to 14.9%. Most of the nonadvanced adenomas, however, were small (<0.5 cm) and had uncertain clinical significance. The detection of advanced adenomas and CRC increased by a much smaller amount.
Overall detection rates of adenomas and cancer may be affected by how thoroughly endoscopists search for flat adenomas and flat cancer. While the phenomenon of flat neoplasms has been appreciated for years in Japan, it has more recently been described in the United States. In a study in which endoscopists used high-resolution white-light endoscopes, flat or nonpolypoid lesions accounted for only 11% of all superficial colon lesions, but these flat or nonpolypoid lesions were about 9.8 times as likely as polypoid lesions to contain cancer (in situ neoplasia or invasive cancer). However, because the definition of flat or nonpolypoid was height less than one-half of the diameter, it is likely that many lesions classified as nonpolypoid in this study would be routinely found and described by U.S. endoscopists as sessile. The existence of very flat or depressed lesions—depressed lesions are very uncommon but highly likely to contain cancer—requires that endoscopists pay more attention to this problem. Flat lesions may play a role in the phenomenon of missed cancers.
A health maintenance organization assessed the impact of ADRs in follow-up after 314,872 colonoscopies performed from 1998 to 2010 by 136 gastroenterologists, each of whom had done at least 300 colonoscopies during that period. The goal was to determine rates of interval CRC, interval advanced CRC, and CRC death and to relate those rates to a gastroenterologist’s ADR. There were 712 interval cancers (155 advanced) and 147 CRC deaths. The risk of interval cancer from lowest-to-highest quintile of ADR was 9.8, 8.6, 8.0, 7.0, and 4.8 per 10,000 person-years of follow-up. The adjusted hazard ratio (HR) for physicians in the highest quintile compared with those in the lowest quintile was 0.52 for any interval CRC, 0.43 for advanced CRC, and 0.38 for fatal CRC. Each 1.0% increase in ADR was associated with a 3% decrease in risk of cancer, although the CI for each quintile was broad. Limitations of the study include the inability to determine which specific feature of ADR led to reduced interval cancer. For example, it is unclear whether it was due to the following:
- Removal of small adenomas that may grow rapidly to become CRC.
- ADR being a surrogate outcome for an endoscopist’s ability to remove adenomas more completely.
- ADR being a surrogate outcome for an endoscopist’s ability to better detect large, flat, serrated lesions.
- Higher ADR leading to recommendations for more frequent postpolypectomy surveillance colonoscopy.
Another limitation is that the harms of a colonoscopy associated with ADR could not be measured.
Nonrandomized controlled trial evidence about colorectal cancer incidence or mortality reduction
Although there is little RCT evidence to assess reduction of CRC incidence or mortality by colonoscopy, some case-control evidence is available. Based on case-control data about sigmoidoscopy, noted above, there was speculation that protection for the right colon might be similar to that for the left colon. A 2009 case-control study of colonoscopy raised questions about whether the impact of colonoscopy on proximal lesions might be different than the impact on distal lesions. Using a province-wide administrative database in Ontario, Canada, investigators compared cases of individuals who were diagnosed with CRC from 1996 to 2001 and had died by 2003. Controls were selected from persons who did not die of CRC. Billing claims were used to assess exposure to previous colonoscopy. The OR for the association between complete colonoscopy and distal lesions was 0.33, suggesting a substantial mortality reduction. For proximal lesions, however, the OR of 0.99 indicated virtually no mortality reduction. However, this study had limited data about whether examinations were complete to the cecum and about bowel prep. Further, many endoscopists were not gastroenterologists.
A case-control study assessed CRC reduction (but not CRC mortality reduction) in the right side versus the left side of the colon. In a population-based study from Germany using data from administrative and medical records, 1,688 patients with CRC were compared with 1,932 participants without CRC, aged 50 years or older. Data were collected about demographics, risk factors, and previous screening examinations. According to colonoscopy records, the cecum was reached 91% of the time. Colonoscopy in the previous 10 years was associated with an OR for any CRC of 0.23, for proximal CRC of 0.44, and for distal CRC of 0.16. While this study did not assess CRC mortality, the results suggested that the magnitude of the right-side versus the left-side difference may be smaller than previously found. It would be extremely useful to assess right-side versus left-side differences in a RCT.
Other case-control data suggest a reduction of CRC incidence on the right side of about 64% compared with about 74% on the left side.
Because there is little RCT evidence and case-control evidence is limited, it is important to consider the degree of mortality reduction from colonoscopy. While a figure of 90% is sometimes cited as the degree of mortality reduction, the question will not be properly answered until the European RCT that has a control group of routine care that involves minimal screening of any kind is completed. Reliable results from colonoscopy RCTs are needed to confirm the studies of FS that suggest a mortality reduction of approximately 50% might occur in the right colon, similar to the demonstrated impact of FS in the left colon. This generalization is limited by a number of factors, including that proximal lesions may have a different pathology (e.g., a serrated appearance and different molecular pathway).
There is a question of whether continued CRC screening after age 75 years is beneficial. An observational cohort study with the Harvard Nurses’ Health Study (NHS) and Health Professionals Follow-Up Study (HPFS) sought to determine whether CRC mortality or morbidity was affected by lower endoscopy (colonoscopy or sigmoidoscopy) screening in individuals older than 75 years, based on a person’s age, comorbidity, family history of CRC, and screening history. Individuals were followed from 1988 to 2016 and had follow-up questionnaires every 2 years, with a 90% response rate. Among over 50,000 individuals, there were 661 CRCs and 323 CRC-related deaths. Screening after age 75 years was associated with reduced CRC incidence (HR, 0.61; 95% CI, 0.51–0.74) and CRC-related mortality (HR, 0.60; 95% CI, 0.46–0.78), regardless of screening history. It is not clear exactly what past screening history meant, however—i.e., the number and results of exams before age 75 years (no polyps; or some polyps indicating increased future risk). The study also assessed the relation of comorbidity to benefit and found that individuals with serious comorbidity—defined as myocardial infarction, stroke, or three comorbidities (high blood pressure, diabetes, and hypercholesterolemia)—did not benefit from screening. Although observational and not an RCT, this study likely provides the best-available data for a scenario in which a formal RCT is unlikely to be performed. It suggests that healthy people older than 75 years may benefit from continued CRC screening and reduction in mortality and morbidity.
Virtual Colonoscopy (Computed Tomographic Colonography )
Virtual colonoscopy (also known as CTC or CT pneumocolon) refers to the examination of computer-generated images of the colon constructed from data obtained from an abdominal CT scan. These images simulate the effect of a conventional colonoscopy. Patients must take laxatives to clean the colon before the procedure, and the colon is insufflated with air (sometimes carbon dioxide) by insertion of a rectal tube just before radiographic examination.
The American College of Radiology Imaging Network group conducted a large, paired-design study, with 2,531 average-risk people (prevalence of polyps or cancer ≥10 mm, 4%; mean age, about 58 years) screened with both CTC and optical colonoscopy (OC). The gold standard was the OC, including repeat OC exams for people with lesions found by CTC but not by OC. Of 109 people with at least one adenoma or cancer 10 mm or larger, 98 (90%) were detected by CTC (referring everyone with a CTC lesion of ≥5 mm). Specificity was 86%, and PPV was 23%. This study raises several concerns, including the following:
- Most, but not all, lesions found by CTC and not by OC were followed up with repeat OC.
- The design did not allow for following patients, thus potentially missing lesions that grow rapidly and would only be seen after follow-up.
- Because the centers conducting the screening were primarily academic centers and the radiologists and endoscopists were carefully trained, the generalizability of the findings is not clear.
- Sixteen percent of participants had an extracolonic finding that required further evaluation.
Unknowns from the study include the following for either OC or CTC:
- The number of detected polyps that would have progressed to invasive cancer.
- The number of people harmed by the screening process.
Another study reported similar sensitivity and specificity in individuals with an increased risk of CRC. In this study, the sensitivity of OC could not be determined because it was done in an unblinded manner. This cross-sectional study suggested that virtual colonoscopy might be an acceptable screening or surveillance test for individuals at high risk of CRC, but it did not address outcome or frequency of testing in those at high risk.
Some studies have assessed how well virtual colonoscopy can detect colorectal polyps without a laxative prep. The question is of great importance for implementation because the laxative prep required by both conventional colonoscopy and virtual colonoscopy is considered a great disadvantage by patients. By tagging feces with iodinated contrast material ingested for several days before the procedure, investigators in one study were able to detect lesions larger than 8 mm with 95% sensitivity and 92% specificity. The particular tagging material used in this study caused about 10% of patients to become nauseated; however, other materials are being assessed.
Another study used a low-fiber diet, orally ingested contrast, and "electronic cleansing," a process that subtracts tagged feces. CTC identified 91% of adenomas 10 mm or larger, but it detected fewer (70%) lesions of at least 8 mm. Patients who received both CTC and OC preferred CTC to OC (290 vs. 175). This study shows that CTC without a laxative prep detects small 1-cm lesions with high sensitivity and is acceptable to patients. Long-term utilization of CTC will depend on several issues, including the frequency of follow-up exams that would be needed to detect smaller lesions that were undetected and may grow over time.
Extracolonic abnormalities are commonly detected with CTC. Fifteen percent of patients in an Australian series of 100 patients, referred for colonography because of symptoms or family history, were found to have extracolonic findings, and 11% of the patients needed further medical workups for renal, splenic, uterine, liver, and gallbladder abnormalities. In another study, 59% of 111 symptomatic patients referred for clinical colonoscopy in a Swedish hospital between June 1998 and September 1999 were found to have moderate or major extracolonic conditions on CTC. CTC was performed immediately before a colonoscopy, and these findings required further evaluation. The extent to which follow-up of these incidental findings benefited patients is unknown.
Sixty-nine percent of 681 asymptomatic patients in Minnesota had extracolonic findings, 10% of which were considered to be highly important by the investigators; these patients required further medical workup. Suspected abnormalities involved kidney (34), chest (22), liver (8), ovary (6), renal or splenic arteries (4), retroperitoneum (3), and pancreas (1); however, the extent to which these findings contribute to benefits or harms is uncertain. Two other studies, one large (N = 2,195) and one small (n = 136) examined the moderate or high importance of extracolonic findings from CTC. The larger study found that 8.6% of patients had an extracolonic finding of at least moderate importance, while 24% of patients in the smaller study required some evaluation for an extracolonic finding. The larger study found nine cancers from these evaluations, at a partial cost (they did not include all costs) of $98.56 per patient initially screened. The smaller study found no important lesions from evaluation, at a cost of $248 per person screened. Both of these estimates of cost are higher than previous studies have found. The extent to which any patients benefited from the detection of extracolonic findings is not clear. Because both of these studies were conducted in academic medical centers, the generalizability to other settings is also not clear. Neither of these studies examined the effect of extracolonic findings on patient anxiety and psychological function.
Technical improvements involving both the interpretation methodology, such as three-dimensional (3-D) imaging, and bowel preparation are under study in many centers. While specificity for detection of polyps is homogeneously high in many studies, sensitivity can vary widely. These variations are attributable to a number of factors including characteristics of the CT scanner and detector, width of collimation, mode of imaging (two dimensional vs. 3-D and/or fly-through), and variability in the expertise of radiologists.
Digital Rectal Examination
A case-control study reported that routine digital rectal examination was not associated with any statistically significant reduction in mortality from distal rectal cancer.
Detection of DNA Mutations in the Stool
The molecular genetic changes that are associated with the development of colorectal adenomas and carcinoma have been well characterized. Advanced techniques have been developed to detect several of these gene mutations that have been shed into the stool. Stool DNA testing was recently assessed in a prospective study of asymptomatic individuals who received colonoscopy, three-card FOBT (Hemoccult II), and stool DNA testing based on a panel of markers assessing 21 mutations. Conducted in a blinded way with prestated hypotheses and analyses, the study found that among 4,404 patients, the DNA panel had a sensitivity for CRC of 51.6% (for all stages of CRC) versus 12.9% for Hemoccult II, while the false-positive rates were 5.6% and 4.8%, respectively.
A next-generation multitargeted stool test combined methylation markers for NDRG4 and BMP3, several KRAS mutations, and a human hemoglobin immunoassay. The markers, each quantitated separately, were combined using an algorithm in a prespecified multivariable analysis. The assay’s sensitivity and specificity were compared with a commercial FIT test (OC FIT-CHEK Polymedco), using colonoscopy as the gold standard. Among 12,776 participants who had colonoscopy screening, were enrolled from 2011 through 2012 at 90 sites in the United States and Canada, and were aged 50 to 84 years (and weighted toward >65 years), 9,989 had fully evaluable results. There were 65 CRC and 757 advanced adenomas or sessile serrated polyps 1 cm or greater. The sensitivity for CRC was 92.3% (60 of 65 CRC) for the multitargeted test and 73.8% for FIT. Sensitivity for advanced lesions was 42.4% for the multitargeted test and 23.8% for FIT. Sensitivity for high-grade dysplasia was 69.2% for the multitarget test and 46.2% for FIT. Sensitivity for serrated sessile polyps 1 cm or greater was 42.4% for the multitargeted test and 5.1% for FIT. Specificities were 86.6% for the multitargeted test and 94.9% for FIT, using nonadvanced or negative colonoscopy results, and were 89.8% and 96.4% for totally negative colonoscopy results. A receiver operating characteristic (ROC) analysis showed that the multitargeted test has higher sensitivity than FIT alone, even when the FIT cutoff is reduced to try to increase sensitivity. A limitation is that there were no data about performance of repeated testing over time and what may be an appropriate testing interval.
Overall, the multitargeted test was more sensitive than FIT for both CRC and advanced precancerous lesions, but the test was less specific. The U.S. Food and Drug Administration approved this multitargeted test for colorectal screening in 2014.
Adherence to Screening
Benefit from CRC screening can only occur if eligible people are screened. There have been problems with screening adherence, particularly for low-income and uninsured people. There are also concerns that some people may adhere less to screening with a colonoscopy than with fecal tests. One well-conducted RCT found that, among an uninsured population, mailed FIT-kit outreach and follow-up reminder phone calls resulted in an adherence rate of 40.7%. Mailed colonoscopy invitations and follow-up phone reminders resulted in a 24.6% adherence rate. The usual-care adherence rate in this trial was 12.1%.
Tailoring Screening to Risk
Benefit of screening might be improved by tailoring the recommended screening test to a person’s degree of CRC risk. For example, if a subgroup of young women were to have a substantially lower risk of proximal neoplasms, then recommending sigmoidoscopy instead of colonoscopy might lead to higher compliance (the U.S. Preventive Services Task Force recommends both procedures without preference, as part of a program of screening those at average risk).
In a study to identify an average-risk group who had a higher versus lower risk of advanced neoplasia (CRC and advanced adenomas) anywhere in the colon, 2,993 individuals having a screening colonoscopy were stratified by age, gender, waist circumference, smoking, and family history (those in high-risk family categories, e.g., Lynch syndrome or adenomatous polyposis coli, were excluded). In a classification system derived in a training set, the risks of advanced neoplasm in four groups were: 1.92%, 4.88%, 9.93%, and 24%. In the two lowest-risk groups, sigmoidoscopy would have detected 51 (73%) of 70 advanced neoplasms. In the independent validation set, results were similar. Whether this system increases overall compliance has yet to be determined.
A similar stratification system based on age, gender, smoking, and family history—and combined with FIT—was tested in Asia to determine whether use of the stratification system plus FIT could detect which people needed colonoscopy. If either the stratification system or FIT was positive, a person was recommended for colonoscopy. Using this strategy, 95% of those with CRC were correctly told to have colonoscopy.