Background

Colorectal cancer (CRC) is the third most common visceral cancer and second deadliest cancer worldwide [1]. In the U.S., CRC is the second most common cancer in men and women combined and the second leading cause of cancer mortality [2]. Despite the effectiveness of screening for CRC, at least 30% of the screen-eligible U.S. population 45–75 years old is not current with screening [2]. While effective in decreasing CRC incidence and mortality [3], colonoscopy is a costly procedure that may burden healthcare systems. Colonoscopy accounts for 55% (nearly $28 B) of the cost of screening for all cancers in the U.S [4]. Further, its invasive nature hinders patient acceptability and presents a risk of rare but serious adverse effects [5]. Globally, screening for CRC remains underutilized, with an overall participation rate of 49.5% in Europe for the fecal immunochemical test (FIT) and 33.2% for the guaiac-based fecal occult blood test (gFOBT) [6].

Annual or biennial fecal blood testing with the FIT is a less costly strategy for both opportunistic and organized, population-based screening. With the ability to quantify the amount of human hemoglobin in feces, FITs have greater sensitivity, comparable specificity, and better uptake than gFOBTs [7]. While gFOBTs are qualitative tests (reported as either positive or negative), FITs produce a quantitative result, measured either in nanograms (ng) of hemoglobin per milliliter (ml) of buffer or in micrograms (ug) of hemoglobin (Hb) per gram (g) of stool but are usually reported qualitatively as positive or negative based on a certain threshold or cutoff, which varies based on the specific FIT, and which may vary from year-to-year based on colonoscopy resources for a particular health care system or country.

Risk factors and risk prediction models can help identify individuals at greatest risk for significant colorectal findings, improving screening resource utilization and efficiency. In the screening setting, the current fecal hemoglobin level may be combined with risk factors to direct use of colonoscopy, improving its yield and efficiency. Among persons who are FIT positive, risk factors may be used to prioritize use of diagnostic colonoscopy among FIT positive individuals. To this end, we conducted a systematic review of the published literature on using risk factors to understand how they affect colonoscopic yield and efficiency in the detection of advanced colorectal neoplasia in the screening setting (in which most FIT results are negative) and in the diagnostic setting (in which all results are positive). The primary study question is: How do risk factors affect the yield of colonoscopy for CRC and advanced, precancerous polyps (together known as “advanced neoplasia”) over FIT alone (1) in the screening setting, and (2) in the diagnostic setting?

Methods

Eligibility criteria

We sought to identify studies that satisfied the following criteria: (i) published in a peer-reviewed journal; (ii) identified risk factors for developing cancer of the colon or rectum (CRC) or advanced neoplasia (the combination of CRC and advanced precancerous polyps); (iii) provided a measure of relative or absolute risk using FIT with one or more phenotypic factors (e.g., age, sex, family history of CRC, body mass index [BMI], or cigarette smoking), and; (iv) applicable to the general population eligible for CRC screening (40 to 75 years old, asymptomatic) or to persons who are FIT positive. Articles were excluded due to: (i) FIT as the only variable (i.e., no covariates were used analytically); (ii) published as an abstract, editorial, or review article; (iii) no risk factors, predictive model or risk score; (iv) a symptomatic population; (v) a study population exclusively younger than age 50; and (vi) the clinical outcome was not advanced neoplasia or CRC.

Information sources and search strategy

Using a combination of subject headings and keywords, we developed a search strategy that includes colorectal cancer, risk factors or risk assessments, FIT, and prediction, model, or score. We performed a comprehensive literature search of Medline (Ovid) and EMBASE (Ovid) from January 2010 to March 2026, limiting to English language literature. The full search strategies for both databases are included in the appendix.

Study identification and screening

We imported the identified references to Covidence, a production tool for systematic reviews, to de-duplicate results and screen titles and abstracts [8]. After the references from the search strategy were downloaded into Covidence, two authors (M.G, M.T., or V.S. and T.I.) independently screened titles and abstracts to exclude publications that were not relevant. Conflicts were resolved through discussion. For articles classified as “maybe/possible”, where a decision could not be made based on title and abstract, we examined the full text. The same authors independently classified all full-text papers within three categories: exclude, include, unclear. Conflicts and unclearly classified papers were discussed, with consensus achieved during face-to-face meetings.

Data extraction

Data were extracted independently by two authors (M.G., M.T., or V.S. and T.I.) using a tailored extraction form. This form included details on methodology including: (i) study objective(s); (ii) sample size; (iii) study design; (iv) study population; (v) study description; (vi) risk factors; (vii) model metrics; (viii) main findings; and (ix) population characteristics, including age, body mass index (BMI) and gender distribution, and prevalence of advanced neoplasia, advanced adenomas and CRC, as defined in each article. In general, advanced adenomas included adenomas 1 cm or larger, or one with villous histology or high-grade dysplasia regardless of size.

Methodological assessment of included studies

Since the present study objective intersects a standard clinical systematic review and a diagnostic test accuracy review, we used criteria and guidelines from the Cochrane GRADE group and modified the QUADAS-2 criteria to create a tool to assess methodological quality specific for this review’s objectives [9]. The grading criteria were: (i) prospective study design (ii) representative population; (iii) accurate measurement of risk factors; (iv) specified brand and FIT cut-off(s); (v) endoscopists blinded to FIT results; (vi) clearly defined outcomes; (vii) independent assessment of each risk factor or risk prediction model to outcome; (viii) whether and how the model was validated; and (ix) (absence of) commercial funding. Each component was rated independently as “Yes”, “No”, “Uncertain” or “Not Available” by two authors (M.G., M.T., or V.S. and T.I.), with discrepant ratings resolved following discussion.

Analysis

For analysis, we summarized each study’s descriptive features, methodological characteristics, and summary-level variables for the study population. We further described each study’s outcomes, metric(s) used to assess those outcomes, and the numerical results for each outcome. For the screening studies, we calculated the yield and efficiency of colonoscopy with FIT alone and with risk factors or risk prediction model for predicting advanced neoplasia, as defined by each study and expressed as the positive predictive value (PPV) (for yield) and number needed to scope (NNS, defined as the number of persons undergoing colonoscopy to identify one with advanced neoplasia) for efficiency. Among the FIT positive studies, we quantified the yield for FIT both based on the threshold for a positive test alone and with risk factors. All data are summarized in tabular format. Due to heterogeneity of study risk-related metrics, FIT thresholds, and risk factors considered, a quantitative summary estimate (i.e., meta-analysis) was not possible.

Results

We screened 1,891 unique titles with or without abstracts, of which 1,860 were excluded due to irrelevance to the study questions. The full text was reviewed for the remaining 31 articles, of which 12 were excluded, leaving 9 screening studies [10,11,12,13,14,15,16,17,18] and 10 diagnostic studies [19,20,21,22,23,24,25,26,27,28] (Fig. 1).

Fig. 1
Fig. 1
Full size image

PRISMA diagram

Screening studies

The nine screening studies are summarized in Tables 1, 2 and 3 and in supplemental Table 1. Table 1 summarizes selected descriptive features of each study. Study objectives and goals were consistent among the studies, all of which are cross-sectional and include asymptomatic average-risk persons generally in the 40-75-year-old range who underwent screening colonoscopy [10,11,12,13,14,15,16,17,18]. Five different FIT types were used at varying cutoffs ranging from 25 ng of hemoglobin per ml of buffer to 100 ng/ml or from 0.2 to 40 µg of hemoglobin per gram of stool, with 5 ng/ml roughly equivalent to 1 µg/g. Risk factors most frequently identified included age, sex, family history of CRC, smoking status – each present in 7 of 9 studies -, BMI (5 studies), and ethanol use (3 studies).

Table 1 Selected descriptive features of the included screening studies
Full size table
Table 2 Selected numerical features of screening studies
Full size table
Table 3 Comparison of yield and efficiency of FIT alone vs. FIT with risk factors in screening studies
Full size table

Descriptive numerical features of the studies

Numerical features of the screening studies are shown in Table 2. Study sample sizes ranged from 948 to 4,332. Mean and median ages were generally homogeneous, ranging from 60 to 64 years, with one exception [15]; sex distribution varied from 30% to 65% female. Prevalence of CRC and advanced adenomas ranged from 0.6% to 1.9% and from 2.9% to 10.9%, respectively. FIT positivity ranged from 3.9% to 36%.

Assessment of study quality

Study quality is shown in supplemental Table 1. While the studies were generally well-designed, there was variation in several quality metrics. Only three studies were prospective in design [11, 12, 15], while all but one study involved only average-risk subjects [10]. Endoscopists were blinded to FIT results in 5 studies [10, 11, 15,16,17]. Some kind of model validation was performed in three studies [10, 15, 18]. All studies measured risk factors accurately, had clearly defined outcomes, pre-specified FIT cutoffs for analysis, and tested for an independent association between risk factors and outcome by using either multivariable analysis or risk stratification. No study declared having competing interests or commercial funding.

Yield and efficiency of FIT alone versus FIT with risk factors

For each study, Table 3 quantifies the yield (PPV) and efficiency (NNS) of FIT alone for detection of advanced neoplasia compared to these metrics when FIT is combined with risk factors or risk models. With the prevalence of advanced neoplasia ranging from 3.7% to 11.8%, the NNS ranged from 8.5 to 23. Both among studies and within studies that used more than one FIT cutoff, the proportion of subjects with a positive FIT varied inversely with the cutoff for a positive test. As such, the yield / PPV ranged from 8% to 48%, with higher yields observed at higher cutoffs (Table 3). The proportion of advanced neoplasia detected within a study also depended on the FIT cutoff, ranging from just under 10% to 51%. Among FIT positive subjects, the NNS to find one subject with advanced neoplasia ranged from as many as 13 to as few as 2.1, with most values in the 2 to 7 range. And as the FIT threshold increased, efficiency improved with higher PPV and lower NNS, but at the cost of missing some proportion of AN.

The same pattern was found when risk factors were considered in addition to FIT: improvements in yield and efficiency were generally modest with the tradeoff of missing some proportion of AN. For example, Omata and colleagues examined FIT thresholds of 25, 50, and 100 ng/ml with respective PPVs of 13%, 15%, and 21%, respective proportions of AN detected of 51%, 35% and 23%, and respective NNS of 7.7, 6.3, and 4.8 [10]. With FIT and risk factors of age, sex, and BMI, the PPV was as low as 3% for a 50-year-old female with BMI of 22 kg/m2 and FIT of < 25 ng/ml and as high as 30% for a 70-year-old male with a BMI of 30 kg/m2 and FIT of 200 ng/ml. When Kallenberg and colleagues added family history to FIT thresholds of 10, 15, and 20 ug/g, the proportions requiring colonoscopy increased, PPV decreased, and NNS increased, but with increases in the proportions of AN detected [13]. Aniwan and colleagues incorporated the Asian Pacific Colorectal Screening (APCS) score, achieving more clinically meaningful risk stratification among subgroups, finding a 6.15-fold risk difference between highest and lowest risk subgroups [11], while He and colleagues found that a modified APCS increased AN sensitivity from 37% for FIT alone to 77% for the combination of FIT and modified APCS [15]. Last and most recently, Wu and colleagues achieved an AUC of 0.87 for fecal hemoglobin level used categorically in addition to risk factors versus 0.796 for binary FIT alone [18].

FIT positive (diagnostic) studies

The ten FIT positive studies are summarized in Tables 4, 5 and 6 and in supplemental Table 2 [19,20,21,22,23,24,25,26,27,28]. Table 4 summarizes descriptive features of each study. Study objectives and goals were consistent among the studies, all of which involved asymptomatic persons generally in the 50-75-year-old range who had a positive screening FIT and underwent diagnostic colonoscopy. All 10 studies are cross sectional [19,20,21,22,23,24,25,26,27,28]. At least three different FITs were used at varying cutoffs ranging from 20 to 160 µg/g. Clinical risk factors identified included age (10 studies), sex (9 studies), fecal hemoglobin value (4 studies), smoking (3 studies), ethanol use (2 studies), among others.

Table 4 Selected descriptive features of FIT positive studies
Full size table
Table 5 Selected numerical features of FIT positive studies
Full size table
Table 6 Comparison of yield and efficiency of FIT alone vs. FIT with risk factors in FIT positive studies
Full size table

Descriptive numerical features of the studies

Numerical features of the studies are shown in Table 5. Study sample size ranged from 743 to 21,530. Mean and median ages ranged from 60 to nearly 72 years. Sex distribution ranged from 38.7% to 52% female. Compared to the screening studies, prevalence of CRC and advanced adenomas were higher, ranging from 1.1% to 9.5% for CRC and from 12.8% to 46.3% for advanced neoplasia, nearly all of which were advanced adenomas.

Assessment of study quality

Study quality assessment is shown in supplemental Table 2. There was greater uniformity in design among these studies as compared with the screening studies. In no studies were endoscopists blinded to FIT results, consistent with clinical practice. In contrast, all studies measured risk factors accurately, had clearly defined outcomes and pre-specified FIT cutoffs, and tested for independent associations between risk factors and outcomes. Four of ten studies performed some kind of model validation [22, 24,25,26]. In none of the studies were commercial funding or competing interests mentioned.

Yield of FIT alone versus FIT with risk factors

Table 6 quantifies the yield / PPV and NNS of FIT positivity alone compared to FIT combined with risk factors for detection of advanced neoplasia. The yield / PPV range from 21% to 65%, with NNS to detect one advanced neoplasm ranging from 1.5 to 4.8. Factors used for risk stratification resulted in modest degrees of discrimination among risk categories. For example, in the study by Kim et al., in which the overall NNS was 4.1, the 5 risk factors resulted in risk quartiles with NNS ranging from 6.5 for quartile 1 with AN risk of 15.5% to 2.4 for quartile 5 with AN risk of 41.5% [21]. Auge and colleagues used age, sex, and quartile of fecal-Hb to go from an overall yield for advanced neoplasia of 46% and NNS of 2.2 to 21.3% and 4.7, respectively, among the lowest risk (women in the lowest quartile) and to nearly 76% and 1.3% (older men in the highest quartile), respectively [19]. In contrast, Cooper and colleagues found little improvement in discrimination over FIT alone with either a logistic model or neural network model [22]. Last and most recently, Bocioaga and colleagues used artificial intelligence modeling to derive and validate a risk score for CRC, achieving an AUC of 0.93 (vs. 0.786 for FIT alone) and identifying a low-risk subgroup representing 36% of the validation cohort in which CRC risk was 0%, as well as a high-risk subgroup of 350 (22.6% of the cohort) that included 96 (86%) of 111 CRCs [26].

Discussion

FIT is the predominant screening test for CRC in the world, especially in countries where colonoscopy is a limited resource. In countries that screen primarily with FIT and have limited capacity for diagnostic colonoscopy, it may be necessary to tailor how FIT is used for screening (e.g., the population age range screened, the cutoff for a positive test) based on how many diagnostic colonoscopies are designated for FIT-positive persons. Further, in countries that offer screening colonoscopy, limitations in capacity, cost, and uptake may result in unintentional “rationing” of this valuable resource. An option to such rationing in these settings is to use risk factors for advanced neoplasia to select persons for FIT vs. colonoscopy screening or, for those with a positive FIT, to prioritize diagnostic colonoscopy, or both strategies, which would improve colonoscopy efficiency and might improve overall uptake of CRC screening. In this systematic review, we describe studies in which risk factors were used in combination with either a screening FIT or a positive FIT to quantify the effect on yield/PPV and NNS. In both screening and diagnostic settings, the effects on the yield and efficiency of colonoscopy for advanced neoplasia when risk factors were used depended on how well the combination of FIT threshold and risk factors delineated different risk groups or improved sensitivity for advanced neoplasia For countries and healthcare systems with limited resources for screening, this information may be useful for understanding the tradeoffs among efficiency, yield, and detection of advanced neoplasms.

Several factors associated with CRC and advanced neoplasia have been identified, resulting in the numerous risk prediction models that estimate current risk for advanced neoplasia [16]. These models may be used to determine who – because of low risk for advanced neoplasia – may be preferably screened non-invasively and who - because of high risk – may be preferably screened with colonoscopy. Models based on polygenic risk scores have been developed and may be useful for determining the age at which screening should start [29]. Although these risk factors and risk prediction models currently are not used to tailor screening, this systematic review indicates that certain factors and risk prediction models (specifically, the well-validated APCS score) may be used to stratify the risk for advanced neoplasia and improve the yield and efficiency of screening colonoscopy and diagnostic colonoscopy among FIT positive persons, although at the cost of missing or deferring detection of some proportion of advanced neoplasia.

In both screening and diagnostic settings, combining risk factors with FIT improved discrimination for advanced neoplasia compared with FIT alone. However, in both sets of studies, the improvement in discrimination and efficiency varied based on FIT cutoff, risk factors considered, and how they were used analytically. Among the 9 screening studies, family history alone and FIT cutoff alone resulted in little improvement in discrimination and efficiency [13]. Combining family history and FIT threshold resulted in nearly identical yields for AN and improved the proportion of AN detected, but at the cost of lower efficiency as measured by the NNS. In contrast, Omata and colleagues used FIT along with age, sex, and BMI, producing a 10-fold difference in PPV when the extreme categories were compared [10].

Both sets of studies illustrate the tradeoff between yield / efficiency and the proportion of advanced neoplasia detected. In general, as efficiency in the use of colonoscopy improved (with fewer being done), the proportion of detected advanced neoplasia decreased – i.e., lesions are missed. This information is most relevant to the screening setting. In the study by Peng and colleagues, foregoing FIT in the low-risk group saved 96 (20.9%) colonoscopies but missed nearly 15% of advanced neoplasms [16]. Whether the tradeoff is worthwhile is a judgment, with missing an advanced adenoma less onerous than missing a cancer, particularly if FIT is administered on an annual basis, in which case FIT may become positive as advanced adenomas increase in size. Population-based studies suggest that the annual rate of malignant transformation of advanced adenomas ranges from 3.5% to 5.5% per year [30]. Thus, most of them will remain benign over the short term, with opportunities for detection through programmatic screening.

Among the FIT positive studies, the study by Kim and colleagues showed that foregoing colonoscopy in persons with the lowest quartile of FIT positivity reduced colonoscopy volume by 25% but missed 6–16% of advanced neoplasms [21]. Despite little improvement in overall discrimination by adding models to FIT, the study by Cooper and colleagues showed that at the higher FIT threshold, the logistic model detected 17 additional advanced adenomas for 13 additional colonoscopies and the neural network model detected 30 advanced adenomas for 24 additional colonoscopies [22]. A limitation of these studies is that only one study separated CRC from advanced adenomas among missed AN. Subsequent studies need to provide more detailed information on the advanced lesions that would have gone undetected had FIT testing been used for screening or if diagnostic colonoscopy had been deferred in low risk FIT positive persons.

Several studies have shown higher levels of fecal occult blood with greater degrees of neoplasia (i.e., no neoplasia, non-advanced neoplasia, advanced adenomas, and adenocarcinoma), indicating that using FIT quantitatively would improve efficiency and could be used to prioritize FIT positive persons for diagnostic colonoscopy [19, 31,32,33]. A recent study by Kang and colleagues showed that increasing quantitative FIT values alone correlated with the PPV for both CRC and advanced adenoma [34]. Further, among persons who are FIT negative, the level of fecal hemoglobin may be used to tailor the interval for retesting [35]. This review indicates that using FIT qualitatively or quantitatively with risk factors improves the yield / efficiency of colonoscopy in both screening and diagnostic settings, the latter exemplified by the studies of Wu and colleagues [18] and Bocioaga and colleagues [26], respectively.

For persons who are FIT positive, timely diagnostic colonoscopy is the standard of care. Observational data indicate that delays in diagnostic colonoscopy are associated with higher risk of CRC and advanced stage CRC [36]. A systematic review of 8 observational studies showed that when colonoscopy is performed beyond 9 months from a positive FIT as compared to within one month, the adjusted risk of CRC is higher, as is the risk of advance-stage disease [37]. In settings where resources for diagnostic colonoscopy are limited, a delay may be unavoidable, in which case, risk for AN may be determined by using the risk factors in these studies with the goal of optimizing waiting times for individual patients.

Operationalizing these findings in the screening setting, diagnostic setting or both requires identifying discriminating risk factors most easily and reliably measured (for example, age, sex, BMI, smoking) and deciding how to use them (for example, generating a risk score). For persons at high risk for advanced neoplasia, colonoscopy may be preferentially recommended, while FIT could be recommended for those with low risk. For those at intermediate risk, either FIT or a choice between FIT and colonoscopy could be recommended. In the diagnostic setting, a very high fecal hemoglobin level warrants colonoscopy sooner, while a level just-above-the positivity threshold may be deferred, depending on presence or absence of other risk factors. Independent of risk for advanced neoplasia, we suggest that patient adherence (with appointments, medications, and other preventive interventions) be considered. This very general approach needs to be tailored based on a country’s or health care system’s goals for screening (e.g., cancer detection, resource optimization, other), ease and reliability of risk factor data collection and information access, screening tests offered, and colonoscopy resources, among other factors.

This systematic review has strengths and limitations that require comment. Strengths include a comprehensive search strategy, clear inclusion and exclusion criteria for study selection, an assessment of study quality based on methodological features that are topic-specific, and structured qualitative and quantitative assessments of each study. The primary limitation is the lack of summary-level numerical metrics that would result from combining the studies statistically (i.e., a meta-analysis). While study design and study quality metrics showed minor variation and were generally good, there was heterogeneity in FIT used (both the test itself and the cutoff for positivity), in the risk factors considered, in how they were modeled, and in how results were expressed. These sources of clinical and methodological heterogeneity precluded combining the studies to obtain a single summary measure of effect. Rather, health systems should decide which study or studies is most consistent with their goals for screening and with ability to operationalize a risk-based system for screening, for diagnostic colonoscopy following a positive FIT, or both. Last, we limited this review to studies that included persons 50 years and older, either largely or exclusively. However, a study by Jung and colleagues suggests that combining FIT with the APCS may extend to persons 40–49 years to stratify risk for advanced neoplasia [38], an age group in which CRC incidence is rising worldwide [39].

Conclusions

This systematic review describes the published literature on using risk factors along with FIT for CRC screening or for diagnostic colonoscopy. In both screening and diagnostic settings, the yield of colonoscopy for advanced neoplasia improves modestly when risk factors are used to stratify who is screened initially with FIT versus colonoscopy and who, among FIT positive subjects, are prioritized for diagnostic colonoscopy because of high risk for advanced neoplasia. This information may be useful for understanding the tradeoffs between resource efficiency, yield for advanced neoplasia, and the proportion detected. Further studies are needed to validate the best performing models in this review, to better quantify the tradeoff between efficiency and both frequency and type of missed advanced colorectal lesion, and to explore alternative approaches to risk stratification through the use of phenotypic factors, other non-invasive screening tests (stool- and blood-based), and perhaps genomic factors.