Risk of Colorectal Cancer and Colorectal Cancer Mortality Beginning One Year after a Negative Fecal Occult Blood Test, among Screen-Eligible 76- to 85-Year-Olds

Colorectal cancer incidence among adults ages 80 to 84 is almost 10 times the incidence among those ages 50 to 54 (Surveillance Epidemiology and End Results, RRID:SCR_006902). Although several screening modalities are effective in reducing colorectal cancer incidence and subsequent mortality (1), the risks from endoscopic screening procedures increase with age (2). In addition, the risk of death from non-colorectal cancer causes increases with age, which limits the potential benefits of screening when the overall death rate from other causes is high. Thus, guidelines generally do not recommend routine colorectal cancer screening beyond the age of 75 years (3–6).

The value of continued colorectal cancer screening beyond age 75 is uncertain (7, 8). For patients ages 76 to 85, the United States Preventive Services Task Force (USPSTF) advises that the decision to screen for colorectal cancer “be an individual one, taking into account the patient's overall health and prior screening history (6).” However, there is little empirical data to inform this individual decision process (6, 9).

Stool-based testing is an attractive screening option (10) compared with more invasive screening tests (e.g., colonoscopy; refs. 11, 12), especially for older adults at high risk for complications: no bowel preparation is required, there is minimal risk of adverse events (13), it can be performed at home, and it is low cost. Owing to the curtailing of nonemergency procedures to conserve resources and reduce infection risk stemming from the COVID-19 pandemic, the use of these low-burden stool-based tests has grown rapidly in recent years (14).

Appropriate colorectal cancer screening of older adults requires provider-patient communication of the balance between long-term benefits and short-term risks to enable informed decision-making (8). In its latest recommendations on colorectal cancer screening, USPSTF grades evidence to support colorectal cancer screening among 76 to 85 year olds as a “C,” indicating that there is at least moderate certainty that the net benefit is small and screening should be offered selectively (6). The potential benefit of colorectal cancer screening is determined by a person's life expectancy, combined with their risks of morbidity and mortality from colorectal cancer. Studies demonstrating the benefits of screening by various testing modalities, including randomized trials of repeat guaiac fecal occult blood testing (gFOBT; refs. 15–18), commonly exclude or have a limited representation of adults older than 75 years (9). Long-term cancer risks among patients with normal and abnormal colonoscopies have recently been published (19–21). However, although some studies suggest that individuals with repeat negative fecal immunochemical tests (FIT) have a relatively low risk of advanced colorectal cancer (22, 23), large population-based studies of colorectal cancer incidence and mortality among older adults with previous screening by stool-based tests are sparse. Given the growing use of stool-based tests for colorectal cancer screening (24) and their dominance in screening programs globally (25), such data are crucial for informing when to potentially stop colorectal cancer screening.

We sought to provide estimates of cumulative colorectal cancer incidence and colorectal cancer mortality among older adults (ages 76–85), following a negative fecal occult blood test 1 year earlier, to inform decision-making based on the potential benefits of continuing screening.

Members of three Kaiser Permanente integrated healthcare systems [Northern California (KPNC), Southern California (KPSC), and Washington (KPWA)] and patients receiving primary care at Parkland Health (PH) in Texas were included in this retrospective cohort study. PH is an integrated healthcare system that serves the under- and uninsured residents of Dallas County. These four healthcare systems contribute to the Optimizing Colorectal Cancer Screening PREcision and Outcomes in CommunIty-baSEd Populations (PRECISE) Research Center, part of the National Cancer Institute–funded Population-based Research to Optimize the Screening Process (PROSPR II) consortium (26). Descriptions of colorectal cancer screening processes and other details of these healthcare systems have been published previously (26, 27). Briefly, all four systems include older populations; the Kaiser Permanente members are generally representative of the census demographics of their respective regional populations (28–31). PH is one of the largest public hospital systems in the United States; patients who become Medicare-eligible at age 65 may choose care outside of PH, which decreases the overall proportion of older adults in PH's patient population. All systems have annually mailed FIT outreach to members not up-to-date with screening by another modality. Screening colonoscopy is available at patient or provider request at the three Kaiser Permanente facilities. KPSC has a greater proportion of members screened by colonoscopy and continues its mailed stool kit program through age 79 years; the other systems cease automatic mailings at age 75 years. This study was conducted in accordance with the U.S. Common Rule. A waiver of consent was granted, and the study was approved by the Institutional Review Boards (IRB) at the study sites and University of Washington.

The PRECISE cohort (Supplementary Fig. S1) served as the source population for this study and comprises patients ages 40 to 95, living at least 1 day between January 1, 2010, and December 31, 2019, and enrolled in (KP) or receiving care at (PH) the four contributing health systems (26). Patients were included if they were ages 76 to 85 sometime between January 1, 2010, and December 31, 2019, had at least 5 years of enrollment in their health plan (KP systems) or primary care utilization within the past 5 years (PH), and were otherwise eligible for colorectal cancer screening. This age group includes people older than 75 years (the upper age universally recommended for screening) but younger than 86 years (when cessation is strongly recommended); it is the 76- to 85-year-old population for which USPSTF recommendations note uncertainty in the benefits of screening and defer to shared decision-making between patients and clinicians regarding screening uptake (32).

The index date for the initiation of follow-up for colorectal cancer incidence and mortality was defined as 365 days after a negative FIT or gFOBT result, that is, the time at which the person might consider screening again, given that stool-based screening is recommended annually (6). Only the earliest index date per person was selected (i.e., in cases where an individual continued to be tested by FIT and had a sequence of negative results after age 76). In the overall PRECISE cohort (ages 40–95), 1.6% of stool-based tests were gFOBT and 98.4% were FIT; for this analysis, all stool-based tests were combined and referred to collectively as “FIT.”

A patient was considered “screen-eligible” if they had no history of inflammatory bowel disease (IBD), adenomas, colorectal cancer, colectomy, or proctectomy prior to their index date. Patients who had up-to-date colorectal cancer screening via sigmoidoscopy in the past 5 years or colonoscopy in the past 10 years were also excluded. Such recent testing would have made a patient “up-to-date” with screening, or, in the case of a positive stool-based test, in need of diagnostic colonoscopy (i.e., ineligible for screening at the time of the index date). CT colonography, barium enema, or abdominal CT procedures during the 180 days prior to the index date were also exclusionary procedures, because evidence of colorectal cancer detected by these imaging procedures would make a participant ineligible for screening at the index date. Screening, by definition, occurs in patients without signs or symptoms of cancer; thus, patients with any signs or symptoms associated with colorectal cancer (e.g., abdominal pain, iron deficiency or unspecified anemia, gastrointestinal bleeding or blood in stools, diarrhea, weight loss or underweight, diverticulitis, constipation, abdominal mass, or change in bowel habits) in the 180 days prior to their index date were also excluded.

Data on patient demographics and clinical characteristics, colorectal cancer screening procedures, colorectal cancer diagnoses, and deaths were obtained from administrative and clinical databases, including electronic health records (EHR), at the study sites.

Certain data elements for PRECISE cohort members were also available prior to their cohort entry, including the date and result of the most recent colorectal cancer tests/procedures (e.g., last colonoscopy, stool test, or sigmoidoscopy prior to cohort entry) and colorectal cancer diagnoses. The look-back period for data on prior testing and conditions varied by site: KPWA extended look-back to January 1, 1993; PH extended to January 1, 1995, for colorectal cancer diagnoses and January 1, 2006, for other data; and KPNC and KPSC extended to January 1, 2000, for enrollment, visits, and colorectal cancer tests, and to the full extent of information available for colorectal cancer diagnoses and gastrointestinal surgeries.

Colorectal cancer diagnoses were obtained from local and central cancer registries: Seattle Puget-Sound Surveillance, Epidemiology and End Results (SEER) registry for KPWA, Texas Cancer and Parkland Health registries for PH, and healthcare system-based cancer registries at KPNC and KPSC, which report to the California Cancer Registry.

KPWA, KPNC, and KPSC received vital status information from state vital records data as well as a variety of internal sources, such as insurance membership and billing, discharge status on claims, etc. At all three KP sites, the primary source of cause of death data was the corresponding state's vital records department; KPSC also obtained cause of death information from internal hospital discharge codes. PH derived vital status data from the Texas Cancer Registry (for participants who were diagnosed with colorectal cancer during cohort eligibility) and from PH EHR; deaths that occurred outside of Dallas County, or at non-Parkland facilities, or were not reported by payers and/or family members, may not have been documented in the PH EHR. Cause of death data for the PH cohort were available only via the Texas Cancer Registry for participants who died after being diagnosed with colorectal cancer in Texas.

Indication for colonoscopies performed after index date was assigned based on manual chart review and natural language processing, or a modified version of a colonoscopy indication algorithm that incorporated administrative and clinical data (33), such as recent procedures, IBD diagnosis, signs and symptoms, past results of colorectal cancer screening tests, and personal history of colorectal cancer.

Comorbidity burden for each patient in the cohort was quantified using the Charlson Comorbidity Index based on EHR data. All sites used diagnosis and procedure code lists updated for ICD-10 (34). Charlson score was assessed as of the last day of the calendar quarter before the index date, computed based on relevant codes from that day and the 364 days prior, and categorized (0, 1, 2, 3, 4, ≥5) for analysis. Other covariates for descriptive strata include year of age at the index date, sex, and number of negative FITs in the 5 years preceding the index date (1 or ≥ 2), where one is the recent negative FIT that occurred exactly 1 year prior to the index date.

Absolute risks of colorectal cancer incidence and colorectal cancer mortality were estimated at 2, 5, and 8 years after the index date using cumulative incidence functions. For each outcome (incidence and mortality) separately, the occurrence of the outcome of interest and any competing events (i.e., death from another cause) was calculated. The cumulative incidence of each outcome of interest is the product of two estimates: hazard for the outcome and estimate of overall survival (i.e., no competing event). Follow-up began on the index date (i.e., 1 year after a negative FIT result, when repeat screening might be considered) and continued until the earliest occurrence of a colorectal cancer-related outcome of interest (i.e., diagnosis of colorectal cancer for incidence estimates or death from colorectal cancer for mortality estimates), death from a cause other than colorectal cancer, or exit from the PRECISE cohort [i.e., due to administrative end on December 31, 2019, aging out of PRECISE cohort eligibility (age >95 years), disenrollment from the health plan (KPNC, KPSC, KPWA), moving out of Dallas County (PH), no healthcare utilization in 3 years (PH), or moving out of the cancer registry coverage area (KPWA)]. Follow-up was censored at 180 days after a screening colonoscopy; these 180 days were included as a buffer to avoid the exclusion of cancers detected by the procedure when performed as a screening test. A colonoscopy performed for a diagnostic indication was not grounds for censoring, because such a colonoscopy is essential clinical care when indicated, and thus a decision about whether or not to continue colorectal cancer screening would not affect the occurrence of a diagnostic procedure. Sensitivity analyses were conducted to confirm that misclassification of indication (for determination of the censoring point) would have a negligible impact on the estimates. Because complete screening by FIT (i.e., including a diagnostic colonoscopy if indicated by a positive FIT result) is associated with a lower risk of mortality from colorectal cancer (35–37), follow-up for colorectal cancer mortality was also censored at 180 days after a subsequent FIT. Censoring at subsequent screening exams thus approximates the desired outcome: risk of colorectal cancer or colorectal cancer-related death in the absence of screening interventions that may have affected the outcome. Deaths from causes other than colorectal cancer were analyzed as competing events, ensuring that the population mortality estimate was based on individuals who were alive (and thus had the potential to be diagnosed with colorectal cancer). Cumulative incidence and mortality estimates were further computed in subgroups stratified by patient characteristics, including age at the index date, sex, ethnicity and race, comorbidity, and number of recent FITs.

For a small proportion of individuals in the analysis, the index date occurred less than 365 days after PRECISE cohort entry because their negative stool test occurred less than 365 days before PRECISE cohort entry. Because only the most recent stool test result prior to PRECISE cohort entry was collected, information about number of tests prior to the one used to define the index date was missing for these individuals. In the main analysis, these patients were classified as having a history of one FIT. In a sensitivity analysis, we excluded these patients from analysis to calculate colorectal cancer incidence and mortality, stratified by history of FIT, and restricted to only individuals we were certain had complete lookback through the 5 years prior to the index date. An exploratory analysis was also conducted in a population restricted to screen-ineligible older adults to explore the magnitude of difference in colorectal cancer incidence and mortality compared with screen-eligible older adults.

The data used in this manuscript are not publicly available due to IRB restrictions to protect patient privacy and consent. Processes for accessing PROSPR II data are described at: https://healthcaredelivery.cancer.gov/prospr/datashare/.

Characteristics of the 118,269 screen-eligible patients ages 76 to 85 at the index date (i.e., with a negative FIT 1 year prior to the index date) are provided in Table 1. More than half of the study population was 76 years old (meaning that their prior negative FIT occurred at age 75), and those ages 76 to 80 comprised 90.5% of the overall study population. More than half of patients were female (56.7%), and 15.5% identified as being of Hispanic ethnicity. Among non-Hispanic patients, the most frequently reported races were White (63.8%), Asian (11.9%), and Black (7.6%). Other racial groups comprised less than 2% of the total population. Comorbidity burden was varied: 40.8% had no major comorbidities (Charlson score = 0) and 23.3% had a Charlson score of three or more. The three most prevalent comorbid conditions at the index date were diabetes (21.4%), renal disease (21.3%), and peripheral vascular disorder (17.1%). Only 3.8% had been hospitalized (overnight inpatient or institutional stay) in the 6 months prior to the index date. Most of the population (61.2%) had experienced two or more negative FIT tests in the 5 years preceding the index date. The study population characteristics stratified by healthcare system are shown in Supplementary Table S1.

Cumulative incidence of colorectal cancer was 0.23% (95% CI, 0.20%–0.26%) 2 years after the index date and 1.21% (95% CI, 1.13%–1.30%) 8 years after the index date (Table 2). Cumulative colorectal cancer mortality was lower: 0.03% (95% CI, 0.02%–0.04%) 2 years after the index date and 0.33% (95% CI, 0.28%–0.39%) 8 years after the index date (Table 3). Risk of death from other causes (competing risk to colorectal cancer mortality estimates) was 4.81% (95% CI, 4.68%–4.96%) 2 years after the index date and 28.40% (95% CI, 27.95%–28.85%) 8 years after the index date (Supplementary Table S2).

Cumulative colorectal cancer incidence and mortality (Tables 2 and 3, respectively) were similar between men and women (Fig. 1). Colorectal cancer incidence was higher for persons ages 81 to 85 compared with ages 76 to 80 at all three time points after the index date, but confidence intervals overlapped for the 8-year estimates (Table 2). However, colorectal cancer mortality was similar across age groups, within the limits of chance, at each time point (Table 3). No differences were observed by ethnicity and race, although small numbers of outcomes impeded the calculation of estimates for some groups (Table 2). Point estimates for mortality at 8 years after the index date were highest for non-Hispanic Native American/Alaska Native patients and lowest for non-Hispanic Asian (Table 3), but they did not differ beyond chance. Estimates were statistically imprecise for persons identifying as Native American/Alaska Native and Native Hawaiian/Other Pacific Islander (Table 3).

Comorbidity scores lacked a clear association with the cumulative risk of colorectal cancer or colorectal cancer mortality (Tables 2 and 3). Across time points, confidence intervals for incidence and mortality estimates in patients grouped by Charlson score overlapped. In contrast, a strong association was observed when comparing comorbidity to risk of death from non-colorectal cancer causes (Supplementary Table S2); for example, we observed a nearly three-fold higher cumulative mortality from non-colorectal cancer causes at 8 years among those with a Charlson score of ≥5 when compared with those with a score of 0 [58.07% (95% CI, 56.21%–59.88%) vs. 19.33% (95% CI, 18.73%–19.94%), respectively].

Risk estimates were stratified according to number of recent FITs (with negative result) in the 5 years preceding index date and age. According to the eligibility criteria, all participants had one FIT exactly 1 year prior to the index date; therefore, any further FITs occurred between 1 and 5 years prior to the index date. Colorectal cancer incidence and colorectal cancer mortality among all individuals with two or more FITs compared with only one FIT were similar, even when further stratified by age group. A sensitivity analysis limited to patients with a complete recording of FITs in the 5 years preceding index showed similar results. At 2 years, colorectal cancer incidence was 0.28% (95% CI, 0.23%–0.33%) for only one FIT versus 0.20% (95% CI, 0.09%–0.39%) for two or more FITs in the 5 years preceding the index; at 8 years after the index date, the colorectal cancer incidence was 1.26% (95% CI, 1.13%–1.39%) for only one FIT versus 1.14% (95% CI, 0.88%–1.46%) for two or more FITs (Table 2). Colorectal cancer mortality estimates were less precise because of the smaller number of outcomes in the population subset (Table 3).

Risk of colorectal cancer diagnosis and colorectal cancer mortality among screen-eligible adults ages 76 to 85 with a negative FIT result 1 year prior were low—1.21% and 0.23% 8 years after index date, respectively—and did not differ markedly by sex, ethnicity, race, or number of recent FITs. The corresponding risk of death from non-colorectal cancer causes was more than 20 times higher than the cumulative colorectal cancer incidence and more than 100 times higher than colorectal cancer mortality (2 years after the index date). In a recent parallel study of screen-eligible older adults with a negative colonoscopy 10 years prior (38), we observed similarly low estimates of cumulative risk of colorectal cancer and colorectal cancer mortality (1.29% and 0.46% at 8 years, respectively).

These estimates among a screened population are lower than the available SEER estimates for nearly the same age group (75–84 years), among which colorectal cancer incidence is 0.58% (2-year) and 2.28% (8-year) and colorectal cancer mortality is 0.20% (2-year) and 0.78% (8-year; refs. 39, 40). Because this study population was defined on the basis of prior screening history (e.g., limited to individuals with a negative FIT result 1 year ago) and excluded anyone ineligible for screening (e.g., due to symptoms or surveillance after high-risk findings), lower estimates were expected in the study population compared with SEER.

The selection of this study population based on a negative test should lead to fewer prevalent cancers, and thus lower the risk of detection of future cancers, for an undetermined period of time (41). Cancers diagnosed during this study follow-up were those that could potentially be detected and treated if screening were to continue between ages 76 and 85, thereby reducing colorectal cancer mortality. Given the moderate sensitivity of FIT (approximately 79%; ref. 42), some cancers could also potentially be missed by the recent FIT and could have been captured, perhaps at a later stage, at a subsequent screening if continued. Thus, the current cumulative risk estimates include cases that were either (i) missed by a recent FIT or (ii) emerged after a prior FIT. The observed lower risk of colorectal cancer mortality among persons with multiple prior FITs is consistent with prior evidence suggesting repeated FIT may be associated with reduced risk for colorectal cancer mortality (35).

These data expand upon existing studies on post-screening colorectal cancer incidence and mortality by focusing on older adults with prior screening. The present estimates also account for the competing risk of death from other causes, which is especially important in older adults. The competing risk approach implicitly acknowledges that lower life expectancy ultimately affects the cumulative risk of colorectal cancer, which provides a useful context for older adults engaged in shared decision-making conversations with their care providers (43).

However, this study has several limitations. First, the study population had an uneven distribution of index ages in the range of interest (i.e., a high proportion of 76 year-olds). The ages available in this observational dataset are limited because the recommended age of colorectal cancer screening cessation is 75 years and (to a lesser extent) because we selected the earliest eligible index date in the observation period. This skewed age distribution should be considered when applying the results to other settings. However, it also means the motivation for completion of the prior negative FIT was, for most patients in the cohort, likely part of a standard screening program and thus more generalizable to other individuals. Second, vital status and cause of death were likely under-ascertained in the PH data, leading to potential underestimation of colorectal cancer mortality. Third, these estimates do not provide individual-level risk prediction, but rather offer empirical evidence of population-level colorectal cancer risk among screen-eligible older adults with a recent negative FIT. As prior studies suggest, those who elect to undergo cancer screening tend to be more likely to adopt a healthier lifestyle overall (44). Therefore, estimates of colorectal cancer risk, colorectal cancer mortality, and other causes of mortality are likely lower in this study population, and differences may not be wholly attributable to the history of a recent negative test. Finally, the event counts in some race strata were insufficient to produce statistically precise estimates.

The strengths of this study include the use of high-quality long-term data (including prior testing history and comprehensive outcome ascertainment using SEER-quality cancer registries) in a large, diverse, multistate, screen-eligible community-based cohort. Another strength of this analysis is that these estimates account for a critical selection factor that may affect most disease risk estimates in older adults: death from other causes. Calculations of cumulative risk estimates that account for competing risks of death acknowledge that lower life expectancy in older populations ultimately affects the cumulative risk of colorectal cancer.

In summary, these findings of low absolute colorectal cancer risk alongside expected competing risk of death from other causes may help inform decision-making regarding whether and when to continue colorectal cancer screening beyond age 75 among screen-eligible adults with a negative FIT test 1 year ago.

R.A. Ziebell reports grants from NCI during the conduct of the study. A. Kamineni reports grants from NCI during the conduct of the study. A.N. Burnett-Hartman reports grants from NCI at the NIH during the conduct of the study. D.A. Corley reports grants from NCI during the conduct of the study. B.B. Green reports grants from NCI during the conduct of the study; nonfinancial support from National Colorectal Cancer Round Table outside the submitted work. T. Levin reports grants from Freenome, Inc. outside the submitted work. J.E. Schottinger reports grants from NIH PROSPR grant during the conduct of the study. J. Chubak reports grants from NCI during the conduct of the study. No disclosures were reported by the other authors.

R.R. Dalmat: Conceptualization, data curation, formal analysis, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. R.A. Ziebell: Data curation, writing–review and editing. A. Kamineni: Conceptualization, funding acquisition, methodology, writing–review and editing. A.I. Phipps: Conceptualization, supervision, methodology, writing–review and editing. N.S. Weiss: Conceptualization, supervision, methodology, writing–review and editing. E.S. Breslau: Conceptualization, writing–review and editing. A.N. Burnett-Hartman: Conceptualization, validation, writing–review and editing. D.A. Corley: Conceptualization, funding acquisition, writing–review and editing. V.P. Doria-Rose: Conceptualization, writing–review and editing. B.B. Green: Conceptualization, writing–review and editing. E.A. Halm: Conceptualization, funding acquisition, writing–review and editing. T.R. Levin: Conceptualization, writing–review and editing. J.E. Schottinger: Conceptualization, funding acquisition, writing–review and editing. J. Chubak: Conceptualization, supervision, funding acquisition, methodology, writing–review and editing.

This work was supported by the NCI at the NIH (UM1CA222035 to Chubak, Corley, Halm, Kamineni, Schottinger) and Dalmat (T32CA009168). This manuscript was written as part of Population-based Research to Optimize the Screening Process (PROSPR II) consortium. The overall aim of PROSPR II is to conduct multisite, coordinated, and transdisciplinary research to evaluate and improve cervical, colorectal, and lung cancer screening processes. The sites comprising the three PROSPR II Research Centers reflect the diversity of US healthcare delivery system organizations.

The views expressed here are those of the authors only and do not necessarily represent the views of the NCI or the NIH.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.