Objective:

We conducted this systematic review to support the U.S. Preventive Services Task Force in updating its recommendation on screening for colorectal cancer (CRC). Our review addresses the effectiveness of CRC screening, the test accuracy of CRC screening modalities, and the harms of CRC screening.

Data Sources:

We updated our prior systematic review and searched MEDLINE, PubMed, and the Cochrane Central Register of Controlled Trials to locate relevant studies for all key questions, from the end of our prior review through December 4, 2019.

Study Selection:

We reviewed 11,306 newly identified abstracts and 502 articles against the specified inclusion criteria. We carried an additional 126 studies forward from our prior review. Eligible studies included English-language studies conducted in asymptomatic screening populations age 40 years and older at average risk or unselected for risk factors. We evaluated direct visualization screening tests and currently available stool-, serum-, and urine-based screening tests. For effectiveness, we included trials or prospective cohort studies with contemporaneous controls; for test accuracy, we included diagnostic accuracy studies using a colonoscopy or cancer registry reference standard; and for harms, we included trials or observational studies reporting serious adverse events.

Data Analysis:

We conducted dual independent critical appraisal of all included studies and extracted all important study details and outcomes from fair- or good-quality studies. We narratively synthesized results by key question and type of screening test. When appropriate, we used random-effects meta-analyses. We graded the overall strength of evidence as high, moderate, low or insufficient based on criteria adapted from the EPC Program.

Results:

Effectiveness. We included 33 unique fair- to good-quality studies that assessed the effectiveness or comparative effectiveness of screening on CRC incidence and mortality. Based on four RCTs (n=458,002), a one- or two-time FS was consistently associated with a decrease in CRC incidence (IRR 0.78 (95% CI, 0.74 to 0.83) and CRC-specific mortality (IRR 0.74 (95% CI, 0.68 to 0.80) compared with no screening at 11 to 17 years of followup. Based on five RCTs (n=419,966), biennial screening with Hemoccult II was associated with a reduction of CRC-specific mortality compared with no screening after two to nine rounds of screening at 11 to 30 years of followup (RR 0.91 [95% CI, 0.84 to 0.98] at 19.5 years; RR 0.78 [95% CI, 0.65 to 0.93] at 30 years). Two prospective observational studies evaluated screening colonoscopy on CRC incidence or mortality. In one study (n=88,902), after 24 years of followup, the CRC-specific mortality rate was lower in people who self-reported at least one screening colonoscopy compared with those who had never had a screening colonoscopy (adjusted HR, 0.32 [95% CI, 0.24 to 0.45]). Results were no longer statistically significant after 5 years in people with a first-degree relative with CRC, as opposed to a sustained association beyond 5 years in people without a family history. Another study (n=348,025) with much shorter followup found that people ages 70 to 74 years who underwent a screening colonoscopy had a lower 8-year standardized risk for CRC (−0.42 percent; 95% CI, −0.24 to −0.63) than those who did not undergo the test. The magnitude of benefit was lower and no longer statistically significant for people ages 75 to 79 years, and this study did not report any mortality outcomes. One prospective study (n=5,417,699) evaluating a national FIT screening program found that one to three rounds of screening with a biennial FIT were associated with lower CRC mortality than no screening (adj RR 0.90, 95% CI, 0.84, 0.95). While 3 Hemoccult II studies include adults under age 50 years, none of these studies conducted subgroup analyses in adults who initiated screening before age 50.

Although we included 21 studies comparing different screening tests in average-risk populations, most of the studies were not true comparative effectiveness studies. Because most of these studies are limited to the evaluation of a single round of screening, report a low CRC yield (number of cancers detected), and do not report interval cancers, they do not provide robust direct evidence of comparative benefit on CRC incidence or mortality outcomes. Several ongoing comparative effectiveness trials that are powered to detect a difference in CRC incidence and/or mortality have not yet reported outcomes.

Test accuracy. We included 59 fair- to good-quality studies evaluating the one-time test accuracy of various screening tests compared to an adequate reference standard.

Direct visualization tests. Only 4 studies (n=4,821) reported the test accuracy of colonoscopy generalizable to community practice. The sensitivity to detect CRC was imprecise because of the limited number of cancers in these studies; the per-person sensitivity ranged from 0.18 to 1.0 (95% CI range 0.01, 1.0). For the detection of adenomas ≥10mm, the sensitivity ranged from 0.89 to 0.95 (95% CI range, 0.70 to 0.99) and the specificity from one study was 0.89 (95% CI, 0.86 to 0.91). For the detection of adenomas ≥6mm, the sensitivity ranged from 0.75 to 0.93 (95% CI range, 0.63 to 0.96) and the specificity was 0.94 (95% CI, 0.92 to 0.96) from one study. Based on 7 studies (n=5,328) of computed tomographic colonography (CTC) with bowel preparation, the per-person sensitivity to detect CRC was again imprecise and the per-person sensitivity ranged from 0.86 to 1.0 (95% CI range, 0.21 to 1.0). For the detection of adenomas ≥10mm, the sensitivity was 0.89 (95% CI, 0.83 to 0.96) and the specificity was 0.94 (95% CI, 0.89 to 1.0). For the detection of adenomas ≥6mm, the sensitivity was 0.86 (95% CI, 0.78 to 0.95) and the specificity was 0.88 (95% CI, 0.83 to 0.95). Based on two studies (n=920) evaluating screening capsule endoscopy, the sensitivity to detect adenomas 10 mm or larger ranged from 0.92 to 1.0 (95% CI range, 0.70 to 1.0) and specificity ranged from 0.95 to 0.98 (95% CI range, 0.93 to 0.99). For adenomas 6 mm or larger, one study reported sensitivity of 0.91 (95% CI, 0.85 to 0.95) and specificity of 0.83 (95% CI, 0.80 to 0.86). Both studies had a high proportion of incomplete exams.

Stool tests. Based on two studies (n=3,503) of Hemoccult Sensa using colonoscopy as a reference standard, sensitivity to detect CRC ranged from 0.50 to 0.75 (95% CI range, 0.09 to 1.0) and specificity ranged from 0.96 to 0.98 (95% CI range, 0.95 to 0.99). Hemoccult Sensa was not sensitive to detect AA. Based on 13 studies (n=44,597) of OC-Sensor family of FITs using colonoscopy as a reference standard, the sensitivity to detect CRC was 0.74 (95% CI, 0.64 to 0.83; I²=31.6%) and the specificity was 0.94 (95% CI, 0.93 to 0.96; I²=96.6%). For the detection of AA, the sensitivity was 0.23 (95% CI, 0.20 to 0.25; I²=47.4%) and the specificity was specificity = 0.96 (95% CI, 0.95 to 0.97; I²=94.8%). OC-Light (k=4, n=32,424) performed similarly to the OC-Sensor family of FITs. Other FITs were not evaluated for CRC detection in more than a single study using a colonoscopy reference standard. Four studies evaluating FIT test performance found no differences in test performance for persons age <50 years compared with older aged adults. Based on 4 studies (n=12,424) of Cologuard (sDNA-FIT) using colonoscopy as a reference standard, the pooled sensitivity to detect CRC was 0.93 (95% CI, 0.87 to 1.0; I²=0%) and the pooled specificity was 0.85 (95% CI, 0.84 to 0.86; I²=37.7%). For the detection of AA, the pooled sensitivity was 0.43 (95% CI, 0.40 to 0.46; I²=0%) and the pooled specificity was 0.89 (95% CI, 0.86 to 0.92; I²=87.8%).

Serum test. Based on one nested case-control study (n=6845), the sensitivity of Epi proColon to detect CRC was 0.68 (95% CI, 0.53 to 0.80), and the specificity was 0.79 (95% CI, 0.77 to 0.81). For the detection of AA, the sensitivity was 0.22 (95% CI, 0.18 to 0.24), and the specificity was 0.79 (95% CI, 0.76 to 0.82).

Urine test. Based on one small study (n=228) in average and high-risk persons, the sensitivity of PolypDx to detect AA was 0.22 (95% CI, 0.18 to 0.24), and the specificity was 0.79 (95% CI, 0.76 to 0.82).

Harms. We included 131 fair- to good-quality studies for the harms of CRC screening. Serious adverse events from a single screening colonoscopy or colonoscopy in asymptomatic persons are relatively uncommon, with a pooled estimate of 3.1 perforations (k=23) (95% CI, 2.3 to 4.0) and 14.6 major bleeds (k=22) (95% CI, 9.4 to 19.9) per 10,000 procedures. Serious adverse events from a single screening FS are even less common, with a pooled estimate of 0.2 perforations (k=11) (95% CI, 0.1 to 0.4) and 0.5 major bleeds (k=10) (95% CI, 0 to 1.3) per 10,000 procedures. Complication rates are higher in colonoscopy following abnormal stool tests or FS. Nineteen studies found increasing rates of serious adverse events with increasing age, including perforation and bleeding. The pooled estimate of perforations for a single screening CTC (k=7) was 1.3 per 10,000 (95% CI, 0 to 2.9). CTC may also have harms resultant from exposure to low-dose ionizing radiation (range, 0.8 to 5.3 mSv per examination). Approximately 1.3 to 11.4 percent of examinations have extracolonic findings that are potentially important requiring diagnostic followup.

Limitations:

Studies comparing different screening modalities to date do not provide evidence of the relative benefit of different screening programs on CRC incidence or mortality. FIT test accuracy is specific to each FIT or family of FITs. Serum testing is promising but to date has only one prospective study evaluating its screening test accuracy. Overall, we have limited data of effectiveness, test accuracy, and harms by age under 50 years, race/ethnicity, or family history. Few studies of endoscopy harms report rates of adverse events in nonendoscopy comparator arms. It is unclear if detecting extracolonic findings represents a true overall benefit or harm.

Conclusions:

Since the 2016 USPSTF recommendation, there is more evidence on effectiveness and test accuracy of newer stool tests (FIT and sDNA-FIT), and the test accuracy of a serum test FDA approved for use in persons declining colonoscopy, FS, gFOBT, or FIT. We also identified a new metabolomic urine test with only one small study with test accuracy data, thus far limited to detection of adenomas. We also have more data on colonoscopy harms demonstrating higher estimates of major bleeding than previously described in 2016. Currently used screening modalities, including colonoscopy, FS, CTC, and various high-sensitivity stool-based tests, and a serum-based test each have different levels of evidence to support their use, different test performance to detect cancer and precursor lesions, and different risks of harms. Recommendations regarding which screening tests to use, or if there is a hierarchy of preferred screening tests, will depend on the decisionmaker’s criteria for sufficiency of evidence and weighing the net benefit.