A Systematic Review of Repeat Fecal Occult Blood Tests for Colorectal Cancer Screening

Colorectal cancer incidence and mortality has declined in the United States since the late 1980s (1), largely due to increasing uptake of screening (2, 3). Guidelines recommend screening with colonoscopy, sigmoidoscopy, fecal occult blood test with high-sensitivity guaiac (gFOBT), or fecal immunochemical test (FIT) starting at age 50 for average-risk adults (4). gFOBT and FIT (hereafter collectively referred to as “FOBT”) have become increasingly common in population-based screening programs in Europe (5), as well as large U.S. health care systems implementing mailed outreach (6, 7). FOBT also plays a critical role in colorectal cancer screening for underserved or rural populations (8, 9), where access to colonoscopy may be limited (10).

Stool-based screening strategies rely on patients completing regular, on-schedule tests (11–13), and failure to complete repeat exams may compromise effectiveness (14). Most European countries, Canada, and Australia recommend stool-based screening every two years, while annual screening is recommended in the United States and Asian countries (15). Compared with the 80%–85% of participants in randomized trials of screening efficacy completing two or more exams (11–13), repeat FOBT in clinical practice settings may be very low or vary widely (16). Repeat FOBT in clinical practice is also complex because it involves reassessing eligibility, considering recommended intervals (annual vs. biennial), and identifying patients due for screening at each round.

Few have characterized repeat FOBT patterns in real-world settings, particularly in light of the growing number of health care systems transitioning to stool-based screening strategies (17) for population health. To address this gap, we conducted a systematic review of the literature to estimate prevalence of repeat FOBT across diverse health care settings and populations.

We conducted all search methods according to the Preferred Reporting of Systematic Reviews and Meta-Analysis (PRISMA) Statement guidelines (18). With the assistance of a health sciences librarian, we searched MEDLINE (via Ovid; 1997 to September Week 4 2017, in-process and other nonindexed citations September 28, 2017 and Epub ahead of print September 28, 2017, searched September 29, 2017), Embase (via Ovid; 1997 to September Week 4, searched September 29, 2017), and the Cochrane Library (via Wiley; Cochrane Database of Systematic Reviews and Cochrane Central Register of Controlled Trials, Issue 9 of 12 September 2017, searched September 29, 2017) for articles published between 1997 and 2017. General concepts that comprised the search included: colorectal cancer, mass screening, screening program, and patient adherence. We adapted search terms for each database's unique keywords and subject headings; strategies were pretested and refined through an iterative process by screening citations for relevance to our eligibility criteria. Search strategies for each database are listed as Supplementary Material. We also hand searched reference lists from eligible articles and Scopus (via Elsevier) to determine whether eligible articles had been cited by others not identified by our search strategy.

We considered articles eligible if they: (i) were written in English; (ii) reported data from a primary study (i.e., not a review, commentary, or editorial); and (iii); measured repeat FOBT over at least two screening rounds. We focused on studies conducted in average-risk populations (e.g., no personal history of inflammatory bowel disease, colorectal cancer, hereditary syndromes, or polyps/adenomas, no family history of colorectal cancer or polyps/adenomas), for whom guidelines at the time recommended initiating screening with FIT or gFOBT at age 50 years (19). To best characterize repeat FOBT in real-world settings, we excluded trials of screening efficacy or intervention studies requiring informed patient consent. We also excluded studies in which the primary outcome was test performance (i.e., sensitivity and specificity).

We screened articles in a multistep process. First, two authors (A. Sen and B. Watson) independently reviewed the titles and abstracts of all articles identified by the search strategy, assigning a rating of “not eligible” or “potentially eligible” for inclusion. A third author (C.C. Murphy) reviewed the title and abstracts of all “potentially eligible” abstracts. Discrepancies in “potentially eligible” ratings across co-authors occurred in fewer than 5% of all abstracts reviewed; all discrepancies were discussed until consensus was reached. Finally, two authors independently evaluated full-text articles of all “potentially eligible” abstracts.

In cases where eligible articles reported data from the same or overlapping patient cohorts, we selected the most recently published article or the article with the most complete data. For example, we identified three articles of overlapping cohorts in the Kaiser Permanente health care system (6, 20, 21), and we report results from the most recent of the three articles (20).

Using an abstraction form created for this review, two authors (A. Sen and B. Watson) extracted relevant information from all eligible articles, including: study setting, sample size, eligibility criteria, and outcome measures. A third author (C.C. Murphy) was available to resolve any discrepancies between the two sets of extracted data. Discrepancies in coding occurred in <5% of all studies and were adjudicated through discussion until consensus was reached across the three co-authors.

Repeat FOBT and relevant outcomes were reported in a variety of ways (e.g., completion of all screening rounds, completion of subsequent screening rounds) across studies. The considerable heterogeneity between studies (I² = 99%) precluded the use of meta-analysis to aggregate effect sizes of repeat FOBT. Therefore, we used reported numbers to manually calculate repeat FOBT as the: (i) proportion of Round 1 participants who completed repeat FOBT in Round 2; (ii) proportion of patients who completed two, consecutive FOBT; or (iii) proportion of patients who completed FOBT in three or more screening rounds (Table 1). When possible, we excluded from our calculation patients with a positive index test, prior colonoscopy, or prior sigmoidoscopy and who would therefore be ineligible for repeat FOBT.

Using the STROBE checklist (22), two authors (A.G. Singal and C.C. Murphy) assessed completeness of reporting on nine selected aspects of internal and external validity related to representativeness, intervention, outcome ascertainment, follow-up period, and eligibility criteria. Each characteristic was assigned a rating (23, 24) of “Y, reported by authors,” “N, not reported by authors,” or “I, inferred by raters but not explicitly reported by authors.” We resolved any discrepancies in rating by discussion until consensus was reached.

There was considerable heterogeneity between studies (I2), and the wide-ranging prevalence estimates precluded the use of meta-analysis to aggregate effect sizes of repeat FOBT.

Our search strategy identified 6,258 potentially eligible articles, of which we reviewed the full text of 312 (see Supplementary Fig. S1 for PRISMA flow diagram). Common reasons for exclusion included evaluating screening performance or efficacy and requiring patient consent. From the full-text review, we identified 35 articles that met inclusion criteria, representing 27 unique studies. As described above, for the eight articles reporting overlapping cohorts, we selected the most recently published article or the article with the most complete data.

Study characteristics are shown in Table 2. Studies were conducted in Europe (n = 12), United States (n = 8), Asia (n = 2), Australia (n = 3), and Canada (n = 2) and represented a variety of health care systems (59.3% mailed, population-based screening outreach, 18.5% mailed outreach in integrated systems, and 25.9% opportunistic screening). Most studies measured repeat FOBT using government health plan or population registry data (n = 17, 63.0%), while others used electronic health records (n = 9, 33.3%). Only one study (25) relied on patient self-report. Studies examined repeat FOBT over a range of 2 to 5 screening rounds. About half (n = 13, 48.1%) of the studies evaluated repeat FOBT across three or more screening rounds, and the remaining studies (n = 14, 51.9%) evaluated repeat FOBT in only two rounds.

Prevalence of repeat FOBT is described in Table 3. Among those who completed FOBT in Round 1, 24.6%–89.6% [median: 82.0%, interquartile range (IQR): 73.7%–84.6%] completed repeat FOBT in Round 2 (16, 26–43). Repeat FOBT appeared higher in mailed outreach programs (26–30, 32, 33, 36–42, 44–46) compared with opportunistic screening (Supplementary Fig. S2; refs. 16, 35, 36, 43). Specifically, the proportion of Round 1 participants who completed repeat FOBT in Round 2 ranged from 69.1% to 89.6% in studies with mailed outreach, whereas repeat FOBT was less than 50% in studies with opportunistic screening. Notably, two pragmatic, randomized controlled trials (36, 47) compared mailed outreach to opportunistic screening in low-income settings. In both trials, a higher proportion of patients randomized to mailed outreach completed repeat FOBT in Round 2 (82.2% vs. 37.3%; ref. 36) and across all screening rounds (30.8% vs. 2.3%; ref. 47) compared with opportunistic screening. There appeared to be only small differences in repeat FOBT in studies with annual (range 34.5%–89.6%) versus biennial (range 24.6%–88.4%) screening (Supplementary Fig. S3), and in studies of FIT versus gFOBT (Supplementary Fig. S4).

The proportion of patients who completed two, consecutive FOBT varied widely across studies, ranging from 16.4% to 80.0% (median: 46.6%, IQR: 40.5%–50.0%; refs. 20, 26–30, 34, 37, 38, 45, 47). Most studies reported repeat FOBT between 40% and 60%. Notable outliers were studies by Garcia and Janda (both <20% completion) and Wong (>80% completion).

Repeat FOBT across all screening rounds also varied, ranging from 0.8% to 64.1% (median: 39.2%, IQR: 19.7%–49.4%; refs. 20, 25, 29, 30, 34, 40, 45–50). Prevalence generally decreased across screening rounds. For example, Gellad and colleagues (51) reported 42.1%, 26.0%, 17.8%, and 14.1% completed one, two, three, and four tests, respectively, over five rounds of screening. Similarly, Pornet and colleagues (52) identified a greater proportion of never (33.6%) or occasional participants (27.7%)—those who completed no or one test over three screening rounds—than consistent participants (38.8%).

Supplementary Table S1 describes the completeness of reporting of each included study. All or the majority of studies described test type, defined repeat FOBT, and used electronic health records (EHR) or registry data to ascertain the outcome. We identified eight studies (6, 20, 21, 25, 35, 37, 38, 40, 41, 47, 51) that did not report type of FOBT, and the study by Bae and colleagues (25) assessed repeat FOBT by patient self-report. Studies were more variable with respect to reporting the number of patients eligible in each screening round or the number who were lost to follow-up, were diagnosed with colorectal cancer or died, or received colonoscopy. Although all studies included patients who were age-eligible for screening (i.e., age 50–75 years), fewer studies made an attempt to exclude patients at higher risk (e.g., family history of colorectal cancer). Some studies (25, 30, 41, 49) required patients to complete a brief questionnaire as part of inclusion criteria.

Success of stool-based screening relies on patients completing regular, on-schedule screening, every one to two years. Studies included in our review report a wide range of repeat FOBT—between 14% and 90%—and prevalence generally declined across successive screening rounds. Our synthesis of data across studies highlight two key challenges: (i) ensuring patients initiate and repeat FOBT consistently as part of stool-based screening strategies; and (ii) increasing the already substantial prevalence of repeat FOBT among patients who have previously initiated screening. As such, interventions that maintain consistent participation in FOBT are needed to optimize the effectiveness of this colorectal cancer screening strategy. Our findings also point to a number of areas for future research and the need for more transparent results reporting.

Although tightly controlled screening efficacy trials report up to 85% of trial participants complete two or more tests, we observed varying prevalence of repeat FOBT across real-world settings. The wide variation in repeat FOBT across studies included in our review underscores potential differences in data collection and quality and highlights the need for better summary measures. Reasons for such wide-ranging prevalence estimates may be related to a variety of factors, including test type and frequency, screening delivery, and intensity of reminders for test completion. Most studies included in our review examined repeat FOBT every two years (i.e., biennial screening), but prevalence in these studies did not appear to differ dramatically from studies of annual screening. Studies also used a variety of test types, and differences in patient handling and collection may have contributed to the wide range of prevalence estimates. In randomized trials of gFOBT versus FIT, participation in FIT screening is about 10% higher than for FOBT (53, 54). Some studies suggest three-sample tests deter patients from completing repeat screening and introduce more opportunity for sampling and collection error (55). Only four studies (16, 45, 47, 48) reported using a three-sample test, and prevalence of repeat FOBT in these studies ranged from 0.8% to 49.8% across all screening rounds. Differences in repeat FOBT by test type (FIT vs. gFOBT) also appeared to be small.

We also observed variability in the proportion of patients completing repeat FOBT depending upon how the outcome was defined. For example, when defined as the proportion of Round 1 participants completing FOBT in Round 2, approximately 75% of patients completed repeat screening. Repeat FOBT was much lower when defined as completion across multiple screening rounds—about 45% of patients completed FOBT in two, consecutive rounds. Repeat FOBT appeared even lower when considering patterns over three or more rounds. These differences in outcome suggest two possible phenomena: (i) prior cancer screening experience predicts repeat, on-schedule screening; and (ii) those who initially refuse are unlikely to participate in subsequent rounds. In the context of interventions, the former suggests FOBT participants should be actively engaged to encourage repeat screening, and nonparticipants may instead benefit from an alternate screening test (56). This variability in outcome is also important when comparing results across studies, which used different definitions for repeat FOBT.

Few studies examined correlates of repeat FOBT, and those that did generally included nonmodifiable factors (e.g., age, sex). This is consistent with studies on correlates and predictors of FOBT initiation, in which sociodemographic variables such as younger age, nonwhite race/ethnicity, low socioeconomic status, poor educational attainment, and lack of insurance are negatively associated with screening uptake (57–59). Although demographic factors may help identify a target population in which to promote screening, they do not identify strategies that can be used to modify or change behavior. Repeat FOBT may depend highly on patient behavior. For example, in our review, Duncan and colleagues (49) found greater perceived barriers and lower levels of response efficacy were associated with drop-out from FOBT screening. Others have shown self-efficacy distinguishes patients engaged in consistent, on-schedule screening from those never screened (49, 60).

Repeat FOBT was generally higher in studies of mailed outreach (either in integrated health care systems or population-based programs) compared with studies of opportunistic screening. Our search strategy also identified two pragmatic trials (36, 47) of screening outreach; both demonstrated the effectiveness of mailed FOBT outreach (i.e., test kits with postage-paid return envelope) to increase patient adherence to two or more tests over multiple screening rounds. Other trials not included in our review similarly show mailed FOBT kits increase one-time screening, regardless of patient factors or preferences (8, 61–64). Incorporating elements of mailed outreach may optimize efforts to implement population health and cancer screening programs. Learning from system-level interventions (65) to promote repeat breast (66) and cervical cancer screening, such as tracking screening utilization and reports to primary care providers, may also help achieve comparable adherence for repeat FOBT.

Our findings also underscore the importance of transparent results reporting to facilitate comparison among studies and health care systems. For example, few studies reported the number of persons eligible at each screening round, and confusion surrounding the appropriate denominator can make it difficult to determine prevalence of repeat FOBT and compare prevalence estimates across studies. Others failed to describe the number of patients completing a prior screening test, creating challenges for measuring the true yield of screening programs. Allison and colleagues (67–69) have developed several standards to improve FIT results reporting, including fecal hemoglobin concentration, sample handling, storage, and transport. Adapting these standards, we have proposed a checklist (Table 4) to strengthen reporting of FOBT screening completion, particularly when assessed across multiple screening rounds. Most importantly, studies of repeat FOBT should report the number eligible at each screening round, including those who become ineligible for a repeat test due to colorectal cancer diagnosis, death, move away from health care system or geographic region, prior positive FOBT and/or diagnostic colonoscopy, and prior colonoscopy for some other reason. These standards will allow researchers to compare and contrast the results of published studies and improve translation of results into clinical practice.

We observed considerable heterogeneity between studies (e.g., different countries, health care systems, test type), and the wide-ranging prevalence estimates precluded the use of meta-analysis to aggregate effect sizes of repeat FOBT. Similarly, because few studies examined correlates, it was not feasible to provide summary estimates. We excluded screening intervention trials requiring informed patient consent, and repeat FOBT may differ in intervention versus clinical practice settings. However, recent post hoc analyses (60, 70, 71) of these trials suggest prevalence of repeat screening is similar to what we reported. Furthermore, many of the studies included in our review reflect European or predominantly insured, white American populations, thereby excluding a number of patients at risk of colorectal cancer and among whom screening uptake remains low (e.g., Hispanics, non-Hispanic blacks). Although we have demonstrated that many patients, including those completing an index FOBT, fail to complete repeat screening, these data do not illustrate specific reasons for suboptimal screening.

In summary, adherence to repeat screening is critical to the effectiveness of stool-based tests, but few patients complete regular, on-scheduling testing over multiple screening rounds. Our review of repeat FOBT showed a wide range of repeat FOBT across 27 studies, as well as varying measures and definitions of repeat screening. Understanding reasons for these patterns may identify strategies to promote regular colorectal cancer screening at recommended intervals.

S. Gupta is an advisor for Freenome, Inc., and Guardant Health, Inc. A.G. Singal is a consultant for Exact Sciences. No potential conflicts of interest were disclosed by the other authors.

The content is solely the responsibility of the authors and does not necessarily represent the official views of CPRIT, NIH, or AHRQ. The funding agencies had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; or preparation, review, and approval of the manuscript.

This study was supported by Cancer Prevention Research Institute of Texas under award number PP160075 (to A.G. Singal, C.C. Murphy), the National Center for Advancing Translational Sciences at the NIH under award number KL2TR001103 (to C.C. Murphy), and Agency for Healthcare Research and Quality under award number R24HS022418 (to A.G. Singal, H. Mayo).