We conducted a systematic review of five key questions to assist the U.S. Preventive Services Task Force (USPSTF) in updating its 2002 recommendation for colorectal cancer (CRC) screening in average-risk adults aged 50 years or older using home fecal occult blood testing (FOBT), flexible sigmoidoscopy (FS), FS FOBT, colonoscopy, or double-contrast barium enema (DCBE). Key questions for this updated review primarily focused on evidence gaps from the previous review: 1) the accuracy (one-time test performance characteristics) and potential harms of newer CRC screening tests—fecal immunochemical tests (FIT), high-sensitivity FOBT, fecal DNA testing, and CT colonography (CTC)—as possible substitutes for currently recommended CRC screening modalities; 2) updating of evidence on the impact of CRC screening on mortality and to estimate the accuracy and harms of colonoscopy and FS in the community setting. A concurrent decision analysis done by others addressed screening program performance, and compared the life-years gained using different CRC screening tests, test intervals, and stopping ages.
We conducted five literature searches of MEDLINE and the Cochrane Library through January 2008. We identified 3948 abstracts from these searches and 488 articles identified from literature searches and outside sources, which we reviewed against specified inclusion-exclusion criteria. Articles were also excluded for quality reasons. Two reviewers' assents were required to exclude a study.
One investigator abstracted key elements of all included studies into standardized evidence tables. A second reviewer verified these data. Two investigators critically appraised and quality-rated all studies. Disagreements were resolved by consensus.
We reported quantitative synthesis for results of each key question, where possible, and qualitative synthesis otherwise. We found no new studies of CRC screening that report mortality outcomes; longer-term follow-up of four biennial FOBT screening trials indicates CRC mortality was reduced 13 to 21 percent after 8 to 13 years of screening in two trials, although another two trials did not show mortality benefit until after 15 to18 years of screening. The Cochrane Collaboration's pooled estimate of CRC mortality reduction in all four FOBT trials at last follow-up was 15 percent, using either random or fixed-effect models (RR 0.85, CI: 0.78,0.92). The largest body of evidence to evaluate screening test performance of newer fecal tests in average-risk screening populations is for fecal immunochemical tests (FITs), which cannot be analyzed as a class, but as individual assay types. Specifically, four individual FITs (Magstream/HemeSelect; FlexSure OBT/Hemoccult ICT; OC-Hemodia; Monohaem) have higher sensitivity for CRC (61 to 91 percent) than estimates for nonrehydrated Hemoccult II (25 to 38 percent) from another recent systematic review, with somewhat reduced specificity (91 to 97 percent). Sensitivity for advanced neoplasia or large adenomas is less commonly reported, but ranges between 20 and 67 percent in FITs, which is comparable or superior to the sensitivity for nonrehydrated Hemoccult II. Better detection appears to occur with 2 to 3 days of sample collection. For FITs, however, there is a mismatch between tests with clinical accuracy data and those with FDA approval and current US market availability. Of the four FITs discussed here, FlexSure OBT/Hemoccult ICT is the only FIT that is both FDA approved and on the US market at the time of this article. Fewer acceptable-quality studies evaluate Hemoccult Sensa, and although it appears to improve sensitivity for CRC (64 to 80 percent), it may also lower specificity (87 to 90 percent). Clinical accuracy data on fecal DNA tests is still too limited to support population screening, and there is a mismatch between available clinical studies and commercially available tests. Where test accuracy results do not indicate superior test sensitivity with comparable specificity, determining the trade-offs between sensitivity and specificity of newer tests for fecal CRC screening in a program of CRC screening requires modeling. Published reports on CT colonography (CTC) screening suggest at least comparable sensitivity to colonoscopy for CRC and large adenomas (10 mm or larger). For smaller polyps (6 mm or larger), published data are inconsistent, with some studies suggesting reduced sensitivity or sensitivity, perhaps contingent upon the CT technology used and the individual reader. Published specificity estimates for CTC are consistently high (≥ 96 percent) for large polyps, but appear lower and more variable (80 to 94 percent) for smaller polyps (6 mm or larger). Test performance estimates will be more precise (more than doubling the number of average-risk patients studied with CTC screening) when currently unpublished data from the ACRIN study are made available. Based on currently published studies, as few as 1 in 8 to 1 in 13 of those screened with CTC would be referred for colonoscopy (if the referral threshold is CTC-detected lesions of 10 mm or greater), or, as many as 1 in 3 to 1 in 5 would be referred for colonoscopy (if the referral threshold is CTC-detected lesions of 6 mm or greater). Few procedure-related harms associated with CTC have been reported, although low-dose ionizing radiation is a potential harm. Additionally, extracolonic findings are relatively common (27 to 69 percent have any findings; 4 to 10 percent have findings of high clinical significance that require treatment or diagnostic evaluation; 5 to 27 percent have findings that would likely require investigation and/or further treatment); the net impact of all of these, in terms of added benefit (or harms), is uncertain. In community settings, FS (with or without biopsy to determine colonoscopy referral) has an estimated sensitivity of 58 to 75 percent for CRC in the entire colon (based on small numbers) and an estimated sensitivity of 72 to 86 percent for advanced neoplasia. Variations in these estimates are likely due to differences in examiner skill and the patient's risks for proximal lesions in the unexamined colon. The performance of FS screening will become more clear after results of current randomized controlled trials (RCT) are reported. While colonoscopy remains the most accurate screening test for CRC at a single application, recent CTC studies have confirmed that colonoscopy misses polyps and may also miss CRC. Colonoscopy also presents a higher risk for harms than other tests. Serious harms from community endoscopies are about ten times more common with colonoscopy (3.1 per 1000 procedures) than with FS (3.4 per 10,000 procedures). The estimates for harms from FS, however, have much wider confidence intervals.
We reviewed the accuracy or harms of a CRC screening test in a single application for each question in this systematic review. The USPSTF commissioned a simultaneous decision analysis comparing different CRC screening programs that addressed repeated screening. Other topics beyond the scope of this review include barium enema for CRC screening, the adherence or acceptability of various CRC screening methods, methods to improve CRC screening rates, and cost-effectiveness.
Based on currently available evidence, refinements in current CRC screening recommendations to add some fecal tests appear warranted. Given potential harms and variation in test accuracy, emphasis on quality standards for implementation of recommended operator-dependent CRC screening tests also appears prudent. Re-evaluation may be appropriate once ongoing RCTs, particularly evaluating CTC, but also evaluating FS and fecal DNA, report their results. Screening for CRC has a rapidly evolving science base, such that guidance may be expected to change as additional research becomes available.
