Colorectal cancer (CRC) is the third leading cause of cancer in the United States [1]. Screening for CRC is effective in preventing cancer deaths as it can lead to detection of early-stage, potentially curable CRC [2, 3]. Colonoscopy and fecal immunochemical testing (FIT) are recommended methods for screening [3, 4]. In addition to CRC detection, colonoscopy serves as a colon cancer prevention test by detection and removal of advanced adenomas (AAs), which are precursors to CRC. In contrast, FIT has good performance for detection of CRCs but not detection of AAs [5]. The estimated sensitivity of AA detection by FIT ranges is approximately 20–30% [4].

While colonoscopy is the gold standard for CRC screening and prevention in the US, use of non-invasive tests like FIT play an important role in expanding CRC screening. This is particularly true in the wake of the COVID-19 pandemic, which led to temporary cessation of outpatient colonoscopy procedures and widespread disruption in CRC screening and diagnoses [6,7,8,9]. To mitigate this disruption, many centers actively expanded access to FIT as primary CRC screening [10]. At our center, FIT has been routinely offered to expand colorectal screening even prior to the COVID-19 pandemic, in part to offset capacity constraints for screening colonoscopy.

As FIT is less effective in detecting AAs, increased FIT administration may lead to an increased number of undetected AAs which may develop into CRCs in the future, thereby paradoxically increasing CRC diagnoses. We determined the potential impact of FIT expansion on missed AAs and future CRC diagnoses in the patient population in an urban, tertiary-care, safety-net hospital.

Methods

Data Source and Study Inclusion/Exclusion Criteria

We created a database of outpatient colonoscopies performed at Boston Medical Center, Boston, MA from 2017 to 2019 by combining data from reimbursement records (Visiquate Inc, Santa Rosa, CA), endoscopy procedural data (Provation, Minneapolis, MN), and pathology reports [11]. We first used Current Procedural Terminology (CPT) codes recorded for reimbursement to exclude colonoscopies of decompression, hemorrhoidal band ligation, endoscopic ultrasound, stent placement, balloon dilation, or foreign body removal. The complete list of included and excluded CPT codes is provided in Supplementary Table 1. We merged this dataset with indications for colonoscopy as reported in Provation. Colonoscopies performed for indications of average-risk CRC screening or positive FIT were included in this study. The demographics of patients between 45 and 79 years attending primary care is shown in supplementary Table 2.

We identified AAs and CRCs by combining data on polyp size from Provation and polyp histology from pathology reports. AA was defined as an adenoma ≥ 10 mm in size, adenoma with tubulovillous/villous histology, adenoma with high-grade dysplasia, per the recommendations of the US Multi-Society Task Force (USMSTF) on Colorectal Cancer [12]. We used reporting systems within Provation to determine individuals who had polyps ≥ 10 mm. Using a natural language search we identified pathology reports with the key words of “tubulovillous,” “villous,” “dysplasia,” “dysplastic,” “carcinoma,” “cancer,” “adenocarcinoma,” and “high-grade” to identify individuals with AA or CRC. These terms were chosen as they are pre-selected by consensus by our pathology colleagues for standardized reporting of colon pathology at our institution. Each pathology record with the above words was reviewed by one of the authors to exclude pathology that did not meet AA criteria. (For example, those that said “negative for adenocarcinoma”).

To validate our database creation strategy and the accuracy of AA and CRCs diagnoses, we reviewed charts of all individuals who underwent colonoscopy for the first 2 months of 2017–2019. Our strategy correctly identified AA vs ‘not AA’ 99.5% of the time and CRC at all times. Finally, as the rate of adenoma to carcinoma progression has been described in individuals > 55 years old, and we had very few individuals > 80 years old, we restricted the analysis to individuals between the ages of 55 and 80 [13].

Statistical Analysis

Expected Number of Yearly Colonoscopies and CRC Screening Tests

The expected number of colonoscopies performed per year by age and indication was determined by averaging the data from 2017 to 2019. We determined the expected number of individuals undergoing CRC screening per year by adding the expected number of colonoscopies performed for CRC screening and the expected number of FIT tests performed per year. To determine the expected number of FIT tests performed per year, we assumed that 7% of individuals undergoing FIT have a positive test result [5], and all FIT-positive individuals are linked to a subsequent colonoscopy. Our center has a robust FIT program, which utilizes patient navigators to ensure return of FIT and linkage of positive FIT to colonoscopy. In the study period, 88.5% individuals with positive FIT either received a colonoscopy or had a documented reason for not receiving one through our center (including external colonoscopy or patient declined).

Estimation of Expected Number of AAs Detected per Year by Age and Gender

AA rate varies by age and gender and is higher in males and with increasing age [14]. We stratified our population by gender and previously described age groups (55–59 years, 60–64 years, 65–69 years, 70–74 years, and 75–79 years) [13]. We used Poisson regression modeling to estimate the frequency of AAs per year by age group using data from 2017 to 2019, assuming average outpatient volume and proportion of screening colonoscopies.

Estimation of Missed AAs with FIT Expansion

As the current estimated use of FIT as the primary method for CRC screening at our center is ~ 50%, we created two models with FIT expansion to 75% and 100% of the population > 55 years old eligible for CRC screening. We assumed that the expected number of individuals undergoing CRC screening tests per year is constant. We estimated AAs detected in these two models using age- and gender-specific AA detection rates in individuals undergoing screening colonoscopies and colonoscopies for positive FIT. The calculations of AA detection for positive FIT are based on the reported performance of single-application FIT. (The effect of serial annual FIT over 5 years on AA detection rate has not been reported). To estimate the number of missed AAs, we subtracted the estimated AAs detected per year under the FIT expansion models from the estimated AAs detected per year under the Poisson model.

Estimate of Progression of Missed AAs to CRC with Expanded FIT Strategy in Population > 55 Years of Age

We used age and gender specific estimates of the annual progression of AA to CRC as previously determined by Brenner et al. [13]. We calculated the cumulative risk of developing CRC in 5 years in each age and gender specific group as 1 − exp (5*rate of annual progression), conditional on not dying from other causes [13]. We multiplied the cumulative risk of developing CRC in 5 years to the age and gender specific missed adenomas estimate to determine predicted new CRCs in 5 years. The 95% confidence interval was calculated by the formula p̂ = p̂ + 1.96 sq rt [(p̂*(1 − p̂)/n] where p̂ is the point estimate and n is the total number of persons screened.

Statistical analysis was performed using Excel 365 (Redmond, WA) and SAS (version 9.4, Cary, NC).

Results

Baseline Characteristics

We identified 6018 colonoscopies performed in the age group 55–80 years, for CRC screening and follow up for positive FIT between 2017 and 2019. Of these, we estimated that 414 (7%) were performed for follow up of positive FIT. During this period, 4257 (41%) colonoscopies were also performed for the same indications in individuals between the age of 45 and 55 years; these were not included in the analyses. Of these, 97% individuals were between 50 and 55 years. In the study period, colonoscopies were not recommended in CRC screening guidelines for individuals between the age of 45 and 49. We noted that ‘screening for CRC’ was the referring indication or listed as an indication by the endoscopist for many individuals in this age group, when they had a normal colonoscopy for a symptomatic indication such as hematochezia as in practice this would serve as the first colonoscopy for CRC screening. Table 1 shows the AA detection rate by age, gender, and indication.

Table 1 Number of colonoscopies performed between 2017 and 2019 for CRC screening and positive FIT, advanced adenomas detected, and advanced adenoma detection rate by age, gender, and indication
Full size table

Expected Number of Yearly Colonoscopies and FIT Tests

Table 2 shows the expected number of colonoscopies per year by indication, expected number of FIT tests per year, and the expected number of individuals undergoing CRC screening by gender. Based on data from 2017 to 2019, it is estimated that 1785 males and 2086 females will undergo screening for CRC per year. If there is no expansion of FIT testing beyond baseline levels, an estimated 946 males (53%) and 1059 (51%) females will undergo FIT testing. This is lines with the current practice at our center, where FIT is offered as the first line CRC screening test in addition colonoscopy to eligible patients attending primary care clinics. Our FIT program has patient navigators to ensure return of FIT and linkage of positive FIT to colonoscopy. Other CRC screening tests like CT colonoscopy, flexible sigmoidoscopy and multitarget DNA tests are rarely used.

Table 2 Expected number of colonoscopies per year (by indication), expected number of FIT tests per year and the expected number of persons undergoing CRC screening, by gender
Full size table

Estimated AA Detection and Missed AAs in FIT Expansion Models

Table 3 shows an estimate of AAs detected and missed, by gender, if FIT testing is expanded to 75% and 100% of the eligible population. With 75% FIT expansion, there would be 21 and 14 missed AAs in males and females, respectively, which is equivalent to one missed AA each for every 85 males and every 153 females screened. With 100% FIT expansion, there would be 46 and 29 missed AAs in males and females, respectively, equivalent to one missed AA each for every 39 males and 72 females screened.

Table 3 Estimate of AAs detected and missed, by gender, if FIT testing is expanded to 75% and 100% of the eligible population
Full size table

Expected Number of CRCs in 5 Years Due to Progression of Missed AAs with FIT Expansion

Table 4 shows estimates of the number of additional CRCs that would be detected in 5 years using previously calculated AA to CRC transition rates for age group and gender. With 75% FIT expansion, we estimate an additional 3.5 (95% CI 1.3, 9.5) and 2.2 (95% CI 0.64, 7.6) CRCs in 5 years in males and females, respectively, which is equivalent to one additional CRC diagnosis in 5 years for every 510 males and 948 females screened. With FIT expansion to 100%, we estimate an additional 7.4 (95% CI 3 0.7, 14.9) and 4.2 (95% CI 1.7, 10.5) CRCs in 5 years in males and females, respectively, which is equivalent to one additional CRC diagnosis in 5 years for every 241 males and 497 females screened. These CRC diagnoses are in addition the estimated number of CRCs detected in our population per year without FIT expansion (4.6 out of 1785 males and 4.6 out of 2086 females screened).

Table 4 Estimated annual transition of advanced adenoma and cumulative risk of developing colorectal cancer in 5 years
Full size table

Discussion

FIT is a non-invasive, inexpensive, easy to administer and well-accepted stool recommended and widely used for CRC screening. It has excellent performance in CRC detection but not in detection of AAs [5]. Expansion of ‘one-time’ (and not serial) FIT, to replace colonoscopy as the modality for CRC screening, in theory, will lead to missed AAs which can progress to CRCs. Conversely, expansion of FIT can also lead to increase in CRC screening uptake and thus improvement of CRC outcomes [4, 13]. Understanding this tradeoff is important in designing CRC screening programs. In this observational study, we estimate the number of missed AAs and potential increase in number of CRCs in 5 years, in our patient population, assuming the screening population is relatively constant every year. We also estimate the number of missed AAs and the potential increase in the number of CRCs in 5 years, in our patient population.

Our results suggest that expansion of ‘one-time’ FIT may substantially increase the CRC incidence. If all 3871 individuals eligible for CRC screening receive FIT (100% expansion), we estimated 2–3 times higher number of incident CRC per year in 5 years (additional 7.3 CRCs per year for males and 4.6 CRCs per year for females). Given this projected increase in CRC cases, we will need to continue to further adjust and study our approach to CRC screening, particularly if any further decrease in endoscopic capacity occurs.

The results of this study are specific to our patient population, but the methodology used can be applied to other populations to generate population-specific estimates. Our patient population is a relatively younger, urban, mid-size, safety-net population. We estimated that approximately 50% of the population received FIT as the screening modality prior to the COVID-19 pandemic, which is higher than the average proportion of CRC screening in the US [15]. Each center’s estimate will depend on the population at risk. We stratified our population by well-defined and well-studies categories of race and gender. However, we were unable to stratify further by race/ethnicity due lack of robust prior data on rates of AA and CRC and due to heterogeneity in report of race/ethnicity in our population.

There is limited evidence to suggest how serial FIT testing may impact the detection of AAs. US guidelines recommend annual FIT testing to improve detection of both AAs and CRC but acknowledge that long-term comparative data to other screening modalities is lacking. Most studies assessing FIT test characteristics were for a one-time test, with the gold-standard comparator being colonoscopy and/or clinical follow-up [4]. Incorporating serial FIT data (was it available) into our models would likely result in at least a modest decrease in our estimates of missed AAs, and subsequently 5-year CRC incidence [16]. Patient adherence to continued yearly FIT testing is inconsistent and ranges widely from 1 to 54% [17], which can also contribute to difficulty in estimating CRC outcomes. To what extent this might be offset by real-world limitations of serial FIT testing including annual compliance and linkage to colonoscopy in a timely manner is unknown.

There are several limitations to our study. Our calculations of AA detection for positive FIT are based on the reported performance of single-application FIT, but the effect of serial annual FIT over 5 years on AA detection rate is not known. We used the age- and gender-specific AA detection rates in our individuals undergoing screening colonoscopy and colonoscopy for positive FIT to determine the estimated number of AAs under each model. While specific to our population, the age- and gender-specific AA rates are like those described by Brenner et al. [13]. Age- and gender-specific AA rates in FIT-positive individuals have not been described. Our pooled AA rates appear to be similar to published gender specific estimates [18, 19]. We restricted our population to those over age 55, as the rate of AA to CRC progression has not been described in the population < 55 years. The estimated number of FIT tests performed per year was extrapolated from the assumption that 7% of individuals undergoing FIT test positive [5], and all FIT-positive individuals are linked to colonoscopy. Failure to include individuals with advanced serrated lesions due to the lack of modeling data on rates of progression may have resulted in an underestimation of projected cancer cases. Finally, our study focuses on proportions of FIT and colonoscopy for CRC screening but does not consider the newer multitarget stool DNA tests [20] as they are not commonly used at our center.

In summary, our findings highlight that expanding FIT to the majority of the eligible average-risk screening population may result in a substantial number of missed AAs, with risk of progression to additional CRC diagnoses within 5 years if counteractive measures are not taken. Further studies will be necessary to determine the impact of expanded FIT testing on incident CRC, and to identify the optimal balance of FIT and colonoscopy for CRC screening to reach the highest proportion of the population without increasing missed AAs and progression to CRC.