Abstract
Background:
Screening colonoscopy harms data are limited for adults ages 76–85 years.
Methods:
We conducted a retrospective cohort study of screening colonoscopies vs. fecal immunochemical tests (FIT) and general population matched comparators aged 76–85 within 3 integrated healthcare systems (2010–2019). The primary outcome was death or overnight hospitalization within 30 days. A secondary outcome also included nine harms diagnoses. Adjusted risk estimates and risk differences (RD) were obtained using Poisson regression. Narrow analyses excluded outcomes after the next lower endoscopy or colorectal procedure while broad analyses included them.
Results:
Patients undergoing screening colonoscopy (N=4435) had a higher 10-day cumulative incidence of gastrointestinal bleeding (0.18% [95% CI: 0.09%, 0.35%]) and perforation (0.09% [95% CI: 0.03%, 0.23%]) than FIT (N=17,740) and the general population (N=44,350) in the narrow analysis. Screening colonoscopy patients had a 1.04% (95% CI: 0.74%, 1.34%) risk of death or hospitalization within 30 days in the narrow analysis, similar to FIT (RD = 0% [95% CI: −0.36%, 0.35%]) and the general population (RD=−0.07% [95% CI: −0.39%, 0.25%]). In the broad analysis, risk following colonoscopy was 2.30% (95% CI: 1.85%, 2.75%) with RD=1.13% (95% CI: 0.67%, 1.60%) vs. general population (age 76–80 RD=0.93% [95% CI 0.45%, 1.41%], age 81–85 RD=2.14% [95% CI: 0.74%, 3.54%]). Secondary outcomes followed a similar pattern by age.
Conclusion:
At ages 76–85, screening colonoscopy including downstream procedures are associated with an increased short-term risk of death or hospitalization.
Impact:
Harms data can be combined with benefits data to guide screening colonoscopy decisions among older adults.
Introduction
Cancer screening recommendations are based, in part, upon the balance of potential benefits and harms. Guidelines starting and stopping ages reflect changes in these considerations over the lifespan. For example, the United States Preventive Services Task Force (USPSTF) recommends colorectal cancer (CRC) screening for adults 45–75 years old but only selectively for people 76–85 years old, largely because of concerns regarding the changing balance of benefit vs. risk with increasing age and comorbidity (1). Across studies included in the evidence review for the USPSTF’s 2021 recommendation, the risks of serious bleeding and colonic perforation following a screening colonoscopy (without age stratification) were 14.6 (95% confidence interval [CI]: 9.4, 19.9) per 10 000 procedures and 3.1 (95% CI: 2.3, 4.0) per 10 000 procedures, respectively (2). A recent systematic review reported estimates of the 30-day risk per 10 000 screening colonoscopies ranged from 16.4 to 36.2 for gastrointestinal (GI) bleeding and 7.6 to 8.5 per 10 000 for perforation (3). None of the estimates, however, was presented by age, though the report for the USPSTF noted that older age was associated with increased risk of harms across indications (2).
Data on harms for older patients undergoing screening colonoscopy specifically are limited. Most studies have not jointly stratified risk by age and indication, and few have provided estimates for screening colonoscopies stratified by age. The best evidence to date comes from a Cleveland Clinic study of approximately 7000 screening colonoscopies in patients 75 years and older, which reported a 10-day incidence of complications requiring hospitalization or an emergency department visit of 13.58 per 1000 persons (4). Risk of complications and severity of complications increased with patient age (i.e., comparing persons 76–80, 81–85, and >85). The incidence of perforation was 0.52 per 1000 for patients 76–80 and higher for those aged 81–85 and >85 (1.95 per 1000 and 3.69 per 1000, respectively). However, without a comparator group to estimate the background risk of events in this older population, it is difficult to know how much of the incidence is attributable to screening colonoscopy vs. other health events that would occur independently of colonoscopy screening.
Given the limited research, data on screening-specific harms in older persons are needed to help patients and providers make informed screening decisions. We therefore conducted a retrospective cohort study to estimate the risk of harms in patients ages 76–85 years receiving screening colonoscopy and compared them to two comparator populations.
Methods
Setting
We conducted a retrospective cohort study within three demographically diverse, community-based integrated health care systems: Kaiser Permanente Southern California (KPSC), Kaiser Permanente Northern California (KPNC), and Kaiser Permanente Washington (KPWA). These populations, which include almost 1 of every 30 people in the United States, closely approximate each region’s underlying demographics (5). Each site collected and harmonized data on patient characteristics, cancer diagnoses, health care utilization, and vital status during 2010–2019 from electronic health records, administrative data, and linkages to cancer registries and state vital records (6). Study procedures were approved by institutional review boards at each site and were in accordance with the US Common Rule and the Declaration of Helsinki. The study was conducted under a waiver of consent.
Study population
We identified screen-eligible patients aged 76 to 85 years old with an outpatient screening colonoscopy between January 1, 2010, and October 31, 2019. Our goal was to identify an average-risk screening population. To ensure we had complete screening history information, we restricted the study to patients who had been enrolled in their health care system for at least 9 years before their screening colonoscopy. Patients were screen-eligible if there was no colonoscopy or unspecified lower endoscopy in the prior 9 years, fecal immunochemical test (FIT) or guaiac-based fecal occult blood test (gFOBT) in the prior 9 months, or sigmoidoscopy in the prior 4 years. Screening indication was ascertained by an administrative data algorithm at KPNC and KPSC (7) and by chart review, natural language processing, and an administrative data algorithm at KPWA (7). We excluded colonoscopies that were preceded by a CRC or inflammatory bowel disease (IBD) diagnosis, recent symptoms or procedures, or prior polyps (Supplementary Table S1), as well as those in people missing information on sex or race and ethnicity. We manually reviewed records of screening colonoscopies in which patients had a GI bleed diagnosed on the day of the colonoscopy and excluded those in which the GI bleed preceded the colonoscopy. The screening colonoscopy date served as the index date.
We used two comparator groups to estimate the background risk of outcomes of interest. The first group comprised patients who completed a FIT. We selected the FIT population as a comparator assuming that, because they were also undergoing screening, they might be similarly healthy (with respect to comorbidities) as patients undergoing screening colonoscopy. We imposed the same restrictions on patients undergoing FIT as those receiving a screening colonoscopy (Supplementary Table S1). We also manually reviewed the records of patients who had a GI bleed diagnosis on the day of the FIT result and excluded those in whom the GI bleed preceded the FIT. The final sample of FIT results was then matched 4:1 to each screening colonoscopy on health care system, age, and year. At KPWA, years 2010–2012 were combined as stool-based screening switched from gFOBT to FIT during this time period and there were relatively fewer FITs during that period. The FIT result date served as the index date, and a patient could have multiple FIT results.
Our second comparison group consisted of screen-eligible patients, sampled quarterly on the 15th of January, April, July, and October of each year 2010–2019, who did not receive a screening or diagnostic colonoscopy on that date (but could have completed a FIT) (Supplementary Table S1). This group, referred to as the general population, was matched 10:1 to screening colonoscopies on health care system, age, and year; the quarterly sampling date served as the index date. We used this group as a comparator in case the FIT group was enriched with people with undiagnosed CRC symptoms (e.g., a person had a FIT kit at home that they decided to take once they started having symptoms that ultimately included a GI bleed). A patient could be included in the general population multiple times.
Outcomes
Outcomes occurring within 60 days of the index date were identified. The primary composite outcome of interest was death or overnight hospitalization. Our secondary composite outcome included the primary composite plus any of the following individual harms diagnosed at either an inpatient or emergency department visit: colonic perforation, splenic injury, GI bleeding, diverticulitis, acute appendicitis, bacteremia, stroke, myocardial infarction, and anesthesia shock. Outcomes were defined using ICD-9-CM and ICD-10-CM codes.
Covariates
In addition to matching factors, covariates of interest at index date included: sex (based on administrative data), comorbidity based on the Charlson score (8), body mass index (BMI), race and ethnicity (based on administrative data), and calendar quarter. A quarterly Charlson score was computed, and the score at the end of the quarter prior to the index date was used. The calculation of BMI was based on the most recent weight in the prior one year and height up to ten years prior. Matching factors and covariates were compared across the screening colonoscopy and comparison groups.
Statistical analysis
We conducted narrow and broad analyses (defined below) to answer slightly different scientific questions in the full study population and stratified by age 76–80 and 81–85 years.
“Narrow” analysis
The goal of the narrow analysis was to estimate the incidence of harms directly related to the index colonoscopy in the screening colonoscopy group and the background rate of harms in the comparator groups. Therefore, we focused on events occurring on the day of or days after the index date but before any subsequent colorectal procedure that was endoscopic (i.e., sigmoidoscopy, colonoscopy, lower endoscopy not otherwise specified) or surgical (colectomy, proctectomy).
We generated cumulative incidence curves for the primary and secondary composite outcomes and individual harms. Follow-up began on the index date and ended at the earliest of the outcome of interest or censoring at a subsequent colorectal procedure, disenrollment, or 60 days after the index date. For individual harms, death was treated as a competing risk. If the outcome occurred on the same day as a subsequent colorectal procedure, we assumed the procedure occurred first, with the following exceptions: 1) When a GI bleed and endoscopic procedure occurred on the same day, we assumed the endoscopic procedure was performed to investigate the bleed; and 2) When a perforation and surgical procedure occurred on the same day, we assumed the surgical procedure was performed to manage the perforation.
Risk differences and relative risks comparing screening colonoscopy to FIT and the general population were estimated for outcomes within a specific time frame and before any censoring or competing risk. For the primary composite outcome of death or overnight hospitalization, the time frame of interest was 0–30 days (Supplementary Figure S1). For the secondary composite outcome (primary outcome or any individual harms), the time frames of interest for individual harms were as follows: colonic perforation 0–10 days, splenic injury 0–10 days, GI bleeding 0–30 days, diverticulitis 0–30 days, acute appendicitis 0–30 days, bacteremia 0–10 days, stroke and myocardial infarction 0–5 days, and anesthesia shock 0–1 days. We used a log Poisson model adjusted for matching factors and covariates (sex, BMI, Charlson score, calendar quarter, and race and ethnicity) to estimate absolute risk estimates, risk differences, and relative risks using a generalized estimating equations (GEE) approach for 95% CIs to account for multiple observations per person (9).
“Broad” analysis
To describe harms more comprehensively, including those that may have resulted from subsequent procedures prompted by a screening colonoscopy, we conducted a second set of analyses that included events after a subsequent colorectal procedure. The rationale for this analysis was that these events may not have occurred were it not for the index screening colonoscopy. We did not use the FIT comparator group in this analysis, because harms in this group of patients—some of whom would undergo colonoscopy after a positive result—would not reflect the background risk in the group eligible for screening colonoscopy.
For cumulative incidence estimation, follow-up began at the index date and ended at the earliest of the outcome of interest, screening colonoscopy, disenrollment, or end of 60-day follow-up. For individual harms, death was treated as a competing risk. For risk differences and relative risks, the same time frames of interests for each outcome were used as in the narrow analysis, with one exception: Outcomes could occur later than the specified time frame of interest if there was a prior colorectal procedure in that time frame. For example, if a patient had a colectomy 40 days after the index screening colonoscopy and then a GI bleed 5 days after that, the GI bleed would be included as an event in the broad analysis but not the narrow analysis. All analyses were performed using SAS software version 9.4 (SAS Institute Inc., Cary, NC); the Margins macro was used to calculate adjusted risk estimates and risk differences (https://support.sas.com/kb/63/038.html).
Data availability
The data generated in this study are not publicly available due to privacy considerations but may be available upon reasonable request from the corresponding author, pending execution of data sharing agreements (contact corresponding author).
Results
After implementing inclusion/exclusion criteria and matching (Figure 1), 4435 screening colonoscopies with one per patient, 17 740 matched FITs among 13 381 patients, and 44 350 matched general population index dates among 35 354 patients were included in this analysis. A total of 50 487 unique people comprised the study sample.
Figure 1. Inclusion criteria for retrospective cohort study of screening colonoscopy harms in patients aged 76 to 85 years old.
Colonoscopies were performed at 43 medical centers by providers. Almost all colonoscopies (98%) were performed by a gastroenterologist and providers had a mean annual volume of 563 (interquartile range 376–737) colonoscopies within the Kaiser Permanente health care system.
Groups were comparable on matching factors and most other demographic variables (Table 1). The mean age of the sample was 78 years old, most patients were non-Hispanic white, females comprised slightly more than half the study population, and over half of people had missing data for BMI. High comorbidity burden (score ≥ 3) was similar among colonoscopy screeners and the general population but slightly lower among FIT screeners. Among those with detailed colonoscopy information available (about two-thirds), most screening colonoscopies had documented adequate bowel preparation (88%) and were complete to the cecum (98%). Slightly over half (52%) had samples sent to pathology. Over 93% of observations were censored at the end of the 60-day follow-up period for all groups and outcomes.
Table 1.
Patient characteristics at time of screening colonoscopy and matched screening FIT 1:4 and general population 1:10
| Screening colonoscopy (N=4,435) |
Screening FIT (N=17,740) |
General population (N=44,350) |
||||
|---|---|---|---|---|---|---|
| N | (%) | N | (%) | N | (%) | |
| Kaiser Permanente regionM | ||||||
| Northern California | 1,817 | (41.0) | 7,268 | (41.0) | 18,170 | (41.0) |
| Southern California | 2,469 | (55.7) | 9,876 | (55.7) | 24,690 | (55.7) |
| Washington | 149 | (3.4) | 596 | (3.4) | 1,490 | (3.4) |
| AgeM | ||||||
| 76 | 1,050 | (23.7) | 4,200 | (23.7) | 10,500 | (23.7) |
| 77 | 983 | (22.2) | 3,932 | (22.2) | 9,830 | (22.2) |
| 78 | 766 | (17.3) | 3,064 | (17.3) | 7,660 | (17.3) |
| 79 | 565 | (12.7) | 2,260 | (12.7) | 5,650 | (12.7) |
| 80 | 335 | (7.6) | 1,340 | (7.6) | 3,350 | (7.6) |
| 81 | 246 | (5.5) | 984 | (5.5) | 2,460 | (5.5) |
| 82 | 187 | (4.2) | 748 | (4.2) | 1,870 | (4.2) |
| 83 | 129 | (2.9) | 516 | (2.9) | 1,290 | (2.9) |
| 84 | 108 | (2.4) | 432 | (2.4) | 1,080 | (2.4) |
| 85 | 66 | (1.5) | 264 | (1.5) | 660 | (1.5) |
| YearM | ||||||
| 2010 | 601 | (13.6) | 2,300 | (13.0) | 6,010 | (13.6) |
| 2011 | 552 | (12.4) | 2,146 | (12.1) | 5,520 | (12.4) |
| 2012 | 493 | (11.1) | 2,138 | (12.1) | 4,930 | (11.1) |
| 2013 | 440 | (9.9) | 1,760 | (9.9) | 4,400 | (9.9) |
| 2014 | 327 | (7.4) | 1,308 | (7.4) | 3,270 | (7.4) |
| 2015 | 381 | (8.6) | 1,524 | (8.6) | 3,810 | (8.6) |
| 2016 | 377 | (8.5) | 1,508 | (8.5) | 3,770 | (8.5) |
| 2017 | 391 | (8.8) | 1,564 | (8.8) | 3,910 | (8.8) |
| 2018 | 435 | (9.8) | 1,740 | (9.8) | 4,350 | (9.8) |
| 2019 | 438 | (9.9) | 1,752 | (9.9) | 4,380 | (9.9) |
| Calendar quarter | ||||||
| 1 | 1,096 | (24.7) | 4,491 | (25.3) | 10,712 | (24.2) |
| 2 | 1,141 | (25.7) | 4,868 | (27.4) | 11,345 | (25.6) |
| 3 | 1,180 | (26.6) | 4,801 | (27.1) | 11,178 | (25.2) |
| 4 | 1,018 | (23.0) | 3,580 | (20.2) | 11,115 | (25.1) |
| Sex | ||||||
| Female | 2,323 | (52.4) | 10,028 | (56.5) | 25,516 | (57.5) |
| Male | 2,112 | (47.6) | 7,712 | (43.5) | 18,834 | (42.5) |
| Race/ethnicity | ||||||
| Hispanic | 653 | (14.7) | 2,609 | (14.7) | 7,458 | (16.8) |
| Non-Hispanic Asian | 516 | (11.6) | 1,994 | (11.2) | 4,634 | (10.4) |
| Non-Hispanic Black | 482 | (10.9) | 1,256 | (7.1) | 3,786 | (8.5) |
| Non-Hispanic White | 2,711 | (61.1) | 11,617 | (65.5) | 27,747 | (62.6) |
| Non-Hispanic Other or Multiracial |
73 | (1.6) | 264 | (1.5) | 725 | (1.6) |
| Body mass index (kg/m2) | ||||||
| <25 | 653 | (14.7) | 2,600 | (14.7) | 5,443 | (12.3) |
| 25-<30 | 644 | (14.5) | 2,458 | (13.9) | 5,612 | (12.7) |
| ≥30 | 322 | (7.3) | 1,324 | (7.5) | 3,885 | (8.8) |
| Missing | 2,816 | (63.5) | 11,358 | (64.0) | 29,410 | (66.3) |
| Comorbidity score* | ||||||
| 0 | 1,804 | (40.7) | 8,033 | (45.3) | 18,546 | (41.8) |
| 1 | 888 | (20.0) | 3,417 | (19.3) | 7,784 | (17.6) |
| 2 | 724 | (16.3) | 2,951 | (16.6) | 7,076 | (16.0) |
| 3+ | 1,019 | (23.0) | 3,339 | (18.8) | 10,944 | (24.7) |
FIT: fecal immunochemical test
Matching variable, FIT matching combined years 2010–2012 for KPWA
Charlson comorbidity score calculation used weights from 1992 Deyo publication (8) updated to include International Classification of Disease, 10th revision codes from both inpatient and ambulatory encounters during 365 days prior to index. Admit date was used for events associated with inpatient stays.
Narrow analysis
There were 727 deaths or hospitalizations within 30 days on or after the index date (screening colonoscopy: n=46, FIT: n=176, general population: n=505). At 30 days, the cumulative incidence of the primary composite endpoint was 1.05% (95% CI: 0.78%, 1.39%) among colonoscopy screeners, 1.00% (95% CI: 0.86%, 1.16%) among FIT screeners, and 1.14% (95% CI: 1.04%, 1.24%) among the general population (Figure 2). The cumulative incidence at other timepoints is provided in Supplementary Table S2.
Figure 2.
Cumulative incidence of outcomes among patients 76–85, by screening status, censored at subsequent colorectal procedure (narrow analysis). (A) Primary composite (death or hospitalization); (B) Secondary composite (death, hospitalization, or emergency room or inpatient harm diagnosis)
A total of 49 GI bleeds (screening colonoscopy: n=10, FIT: n=13, general population: n=26) and 4 perforations (all in the screening colonoscopy group) were observed on or within 30 days. GI bleeds and perforations occurred more often after screening colonoscopy than in both comparison groups throughout the follow-up period (Figure 3, Supplementary Table S2). The cumulative incidence of GI bleeds at 10 days was 0.18% (95% CI: 0.09%, 0.35%) for screening colonoscopy, 0.06% (95% CI: 0.03%, 0.10%) for FIT, and 0.02% (95% CI: 0.01%, 0.04%) for the general population and at 30 days was 0.23% (95% CI: 0.12%, 0.41%), 0.07% (95% CI: 0.04%, 0.12%), and 0.06% (95% CI: 0.04%, 0.09%), respectively. For perforations, the cumulative incidence was 0.09% (95% CI: 0.03%, 0.23%) at both 10 and 30 days for screening colonoscopy, and no events were observed through 30 days in either comparison group (Supplementary Table S2). Splenic injury, acute appendicitis, and shock were extremely rare. The 30-day cumulative incidence of death was highest in the general population. The cumulative incidence of the secondary outcome (death, hospitalization, or inpatient/emergency department harm diagnosis) for screening colonoscopy, FIT, and the general population was 1.39% (95 CI: 1.07%, 1.77%), 1.11% (95% CI: 0.97%, 1.28%), and 1.26% (95 CI: 1.15%, 1.36%), respectively, at 30 days. The cumulative incidence of each component of the secondary composite outcome is shown in Supplementary Figure S2.
Figure 3.
Cumulative incidence of individual harms diagnosed during an inpatient or emergency room visit, among patients 76–85, by screening status, censored at subsequent colorectal procedure (narrow analysis)
Based on the Poisson model, the screening colonoscopy group had an adjusted absolute risk of 1.04% (95% CI: 0.74%, 1.34%) for the primary outcome and 1.25% (95 CI: 0.92%, 1.57%) for the secondary outcome, with outcomes more common among older (81–85 years) vs. younger (76–80 years) patients. No differences in adjusted absolute risk for the primary outcome were found between screening colonoscopy and either of the comparison groups overall (risk difference vs. FIT = 0% [95% CI: −0.36%, 0.35%], risk difference vs. general population = −0.07% [95% CI: −0.39%, 0.25%]) (Table 2). Within age strata, risk differences comparing screening colonoscopy and the general population were −0.17% (95% CI: −0.49%, 0.16%) for ages 76–80 and 0.36% (95% CI: −0.63%, 1.35%) for ages 81–85 and compared to FIT were −0.08% (95% CI: −0.44%, 0.28%) for ages 76–80 and 0.33% (95% CI: −0.81%, 1.46%) for ages 81–85.
Table 2.
Risk, risk difference, and relative risk of harms following screening colonoscopy in patients 76–85 compared to FIT and general population
| Primary outcome (death or hospitalization within 30 days) |
Secondary outcome (death or hospitalization within 30 days or harm diagnosis within specific time frame§) |
|||||
|---|---|---|---|---|---|---|
| Risk estimate | Risk difference | Relative risk | Risk estimate | Risk difference | Relative risk | |
| Ages 76–85 | ||||||
| Narrow analysis † | ||||||
| - Screening colonoscopy | 1.04% (0.74%, 1.34%) | ----- | ----- | 1.25% (0.92%, 1.57%) | ----- | ----- |
| - General population | 1.11% (1.01%, 1.22%) | -0.07% (−0.39%, 0.25%) | 0.94 (0.69, 1.27) | 1.17% (1.06%, 1.28%) | 0.08% (−0.26%, 0.43%) | 1.07 (0.81, 1.41) |
| - FIT | 1.05% (0.86%, 1.24%) | -0.00% (−0.36%, 0.35%) | 1.00 (0.71, 1.40) | 1.11% (0.92%, 1.31%) | 0.13% (−0.25%, 0.52%) | 1.12 (0.82, 1.54) |
| Broad analysis ǂ | ||||||
| - Screening colonoscopy | 2.30% (1.85%, 2.75%) | ----- | ----- | 2.48% (2.01%, 2.95%) | ----- | ----- |
| - General population | 1.17% (1.06%, 1.28%) | 1.13% (0.67%, 1.60%) | 1.97 (1.59, 2.45) | 1.22% (1.10%, 1.33%) | 1.26% (0.78%, 1.75%) | 2.04 (1.65, 2.52) |
| Ages 76–80 | ||||||
| Narrow analysis † | ||||||
| - Screening colonoscopy | 0.88% (0.58%, 1.19%) | ----- | ----- | 1.07% (0.74%, 1.41%) | ----- | ----- |
| - General population | 1.05% (0.93%, 1.16%) | -0.17% (−0.49%, 0.16%) | 0.84 (0.59, 1.21) | 1.10% (0.98%, 1.22%) | -0.03% (−0.38%, 0.33%) | 0.98 (0.70, 1.36) |
| - FIT | 0.96% (0.77%, 1.15%) | -0.08% (−0.44%, 0.28%) | 0.92 (0.62, 1.37) | 1.02% (0.82%, 1.23%) | 0.05% (−0.34%, 0.44%) | 1.05 (0.72, 1.52) |
| Broad analysis ǂ | ||||||
| - Screening colonoscopy | 2.03% (1.56%, 2.50%) | ----- | ----- | 2.22% (1.73%, 2.71%) | ----- | ----- |
| - General population | 1.10% (0.98%, 1.22%) | 0.93% (0.45%, 1.41%) | 1.84 (1.43, 2.38) | 1.15% (1.03%, 1.27%) | 1.07% (0.56%, 1.57%) | 1.93 (1.51, 2.47) |
| Ages 81–85 | ||||||
| Narrow analysis † | ||||||
| - Screening colonoscopy | 1.81% (0.86%, 2.77%) | ----- | ----- | 2.08% (1.06%, 3.10%) | ----- | ----- |
| - General population | 1.45% (1.18%, 1.73%) | 0.36% (−0.63%, 1.35%) | 1.25 (0.71, 2.18) | 1.51% (1.22%, 1.79%) | 0.57% (−0.48%, 1.63%) | 1.38 (0.82, 2.33) |
| - FIT | 1.48% (0.89%, 2.08%) | 0.33% (−0.81%, 1.46%) | 1.22 (0.63, 2.38) | 1.55% (0.95%, 2.15%) | 0.53% (−0.67%, 1.72%) | 1.34 (0.71, 2.52) |
| Broad analysis ǂ | ||||||
| - Screening colonoscopy | 3.64% (2.27%, 5.01%) | ----- | ----- | 3.81% (2.41%, 5.21%) | ----- | ----- |
| - General population | 1.50% (1.21%, 1.78%) | 2.14% (0.74%, 3.54%) | 2.43 (1.59, 3.71) | 1.55% (1.26%, 1.84%) | 2.26% (0.83%, 3.69%) | 2.46 (1.63, 3.72) |
FIT: fecal immunochemical test
Adjusted for matching factors (site, age, year), quarter (1, 2, 3 or 4), sex (female, male; other or missing excluded), race and ethnicity (Hispanic, Non-Hispanic Asian, Black, White, or Other/Multiple; missing excluded), BMI (<25, 25-<30, 30+, missing), and comorbidity (0, 1, 2, 3+)
Narrow analysis did not include events after a subsequent colorectal procedure
Broad analysis included events within specific time frame of a subsequent colorectal procedure (30 days for death or hospitalization or harm diagnosis time frame§)
Harm diagnosis (inpatient or emergency department) time frame: colonic perforation 0–10 days, splenic injury 0–10 days, GI bleeding 0–30 days, diverticulitis 0–30 days, acute appendicitis 0–30 days, bacteremia 0–10 days, stroke and myocardial infarction 0–5 days, and anesthesia shock 0–1 day
Broad analysis
In the broad analysis, which included events after a subsequent colorectal procedure, cumulative incidences of the primary and secondary composite outcomes were higher among screening colonoscopy patients than the general population (Figure 4A and 4B).
Figure 4.
Cumulative incidence of outcomes among patients 76–85, by screening status, censored at screening colonoscopy in general population (broad analysis). (A) Primary composite (death or hospitalization); (B) Secondary composite (death, hospitalization, or emergency room or inpatient harm diagnosis)
At 30 days, cumulative incidences for the primary composite outcome were 1.65% (95% CI: 1.30%, 2.05%) for screening colonoscopy and 1.15% (95% CI: 1.06%, 1.26%) for the general population (Supplementary Table S3). The secondary composite outcome cumulative incidences were 1.94% (95% CI: 1.56%, 2.38%) for screening colonoscopy and 1.27% (95% CI: 1.17%, 1.38%) for the general population. The adjusted absolute risk of the primary composite outcome was 2.30% (95% CI: 1.85%, 2.75%) in the screening colonoscopy group, 1.13% (95% CI: 0.67%, 1.60%) higher than the general population (1.17% [95% CI: 1.06%, 1.28%]). The risk difference between screening colonoscopy and the general population was 0.93% (95% CI: 0.45%, 1.41%) for ages 76–80 and 2.14% (0.74%, 3.54%) for ages 81–85 years. A similar pattern was observed for the secondary outcome in both the narrow and broad analysis (Table 2). The broad analysis included an additional 52 secondary composite outcomes with the majority (N=43, 83%) occurring on or after a subsequent surgery (42 partial colectomies, 1 total colectomy). The remaining patients had an inpatient hospitalization after a lower endoscopy procedure (N=6) or lower endoscopy followed by surgery (n=3).
Discussion
In this large study of screening colonoscopy among patients 76–85 years old performed in three integrated health systems by clinicians with high procedure volumes we did not find evidence that the risk of death or hospitalization within 30 days (1.04%) was higher than in the general population or among patients screening with FIT based on events only up until the next colorectal procedure (i.e., narrow analysis). When events occurring after those procedures were included (i.e., broad analyses), patients receiving screening colonoscopy had a higher risk of death and hospitalization than the general population (2.30% vs. 1.17%). Thus, even though screening colonoscopy itself may not substantially increase the risk of hospitalization or death, it can lead to other procedures that may carry further risk of these outcomes. Other colorectal procedures following screening colonoscopy, like colectomy to remove cancer, may be warranted and overall beneficial; however, they still contribute to risks that patients may expect after a screening colonoscopy. Conclusions were similar for the secondary outcome of death, hospitalization, or harms diagnosed at either an inpatient stay or emergency department visit.
Comparison to other studies
Data on harms among patients >75 years old undergoing screening colonoscopy specifically are sparse. Differences in eligibility criteria, length of follow-up, and outcome definitions make comparison across studies challenging. However, in general, we estimated crude cumulative incidences to be lower than or comparable to those in other studies. El Halabi et al. examined outcomes following 7076 screening colonoscopies in patients >75 years old at the Cleveland Clinic (Ohio and Florida) (4). “Adverse events” were those requiring hospitalization or an emergency department visit and classified by medical record review as related to the procedure. The incidence of adverse events within 10 days was 1.36% and was higher at 30 days (shown in a figure but not reported). In contrast, the 10-day crude cumulative incidence of the primary outcome (death or hospitalization) and secondary outcome (death, hospitalization, or emergency department/inpatient diagnoses) in our study were 0.57% and 0.79%.
Causada-Calo et al. studied 7626 screening and surveillance colonoscopies among patients ≥75 years old in Ontario, Canada (10). They reported that 6.8% of patients had complications (unplanned hospital admission or emergency department visit) within 30 days of colonoscopy, with 0.08% experiencing a perforation and 0.9% experiencing bleeding. Their estimate for complications may have been higher than ours because their population was older (median age 80.5 years vs. 78 years in our study).
Rutter et al. previously estimated the risk of adverse events at Group Health, which became Kaiser Permanente Washington (11); however, there was no overlap in years between our two studies. Rutter et al. included patients followed up for a positive FIT and reported slightly higher incidence for outcomes within 30 days than the current study: death (Rutter: 0.13%; current study 0.07% (95% CI: 0.02%, 0.20%)), hemorrhage (Rutter: 0.81%) vs. bleed (current study: 0.23% [95% CI: 0.12%, 0.41%]), perforation (Rutter: 0.17%; current study: 0.09% [95% CI: 0.03%, 0.23%]), hospitalization (Rutter: 2.7%; current study: 1.00% [95% CI: 0.74%, 1.33%]), emergency department/urgent care (Rutter: 3.5%) vs. emergency department/hospitalization/death; current study 1.05% [95% CI: 0.78%, 1.39%]). These differences may stem from different colonoscopy indications, secular trends, the fact that Rutter et al. attempted to focus on first colonoscopies, or that emergency department visits/hospitalizations were not restricted to specific diagnoses in their analysis.
Grossberg et al. conducted a single-center analysis of approximately 2500 patients >75 years old undergoing screening or surveillance colonoscopy (12). The percent with an emergency department visit in the subsequent 7 days was 1.2% for patients aged 76–85 and 1.6% for patients >85 years of age. For hospitalization, these estimates were 0.7% and 1.6% respectively.
Strengths and limitations
Strengths of this analysis include: its large multi-center design, encompassing 43 medical centers across several geographic regions; the community-based nature of the population; the diversity of population demographics, which approximate the regions’ underlying demographics (5); the use of chart abstraction to ensure that GI bleeding on the day of the screening colonoscopy did not occur before the screening procedure; analysis of procedure-related complications in biologically plausible intervals for each complication; and estimation of the excess risks of death and hospitalization compared to persons not screened by colonoscopy. The USPSTF-commissioned review on CRC screening noted that not having a comparison group was a limitation of other studies (2). Failing to “subtract out” the background risk of death and hospitalization in this older population could substantially overestimate the risks from screening colonoscopy.
Several important limitations are also worth noting. First, the composite outcome of death and hospitalization (our primary endpoint) does not capture all screening-related harms that are relevant to patients, providers, and health systems. We tried to address this with our secondary outcome (which included specific diagnoses that required inpatient or emergency department care) and in analyses of specific harms. However, even with a large sample size, we could not precisely estimate the risk of individual harms. While this is a limitation, it also reflects how rare such outcomes are.
A second limitation is confounding by health status: If patients who screened were healthier than in the general population, we may have underestimated the risk of death and hospitalization associated with screening. We tried to address this concern by comparing screeners to patients completing FIT screening, assuming that they would be more similar to patients undergoing colonoscopy screening than the general population. However, the FIT comparators had a lower comorbidity burden than colonoscopy screeners. Thus, to the extent that diagnostic or surveillance colonoscopies were inadvertently included in our study, and if they carry a higher risk of adverse outcomes, the results might have overestimated risks associated with screening colonoscopy.
A third limitation was that our dataset did not include information on medications (e.g., anti-thrombotic or anti-coagulant drugs) that might have increased the risk of harms, so we were unable to stratify results by this factor. Finally, results may vary across settings to the extent that patient characteristics (e.g., age, comorbidity, medication use) and procedures (e.g., performing providers, polypectomy) differ.
A fourth limitation is that only people who underwent colonoscopy were included in this study. Those who started but could not complete the bowel preparation or had contraindications on the day they were scheduled for a procedure were not included. Thus, any harms in this subgroup were not identified.
Fifth, it is worth noting that the majority of patients in this study were 76–80 years old. While patterns of results were similar across the two age groups examined (76–80 and 81–85), our results suggest that incidence might be higher among older patients. Further work focused on this older group is warranted, especially with a large enough sample size to stratify by comorbidity status.
Finally, we did not evaluate the potential benefits of screening. Ideally, decisions about screening should be based on a balance of potential benefits and harms.
Summary
Our results provide needed information for patients and providers considering whether to undergo or recommend CRC screening for adults aged 76–85. Absolute estimates of risk, as well as comparisons with people not undergoing colonoscopy screening, provide evidence to complement studies of screening benefits in older persons.
Supplementary Material
Acknowledgments:
Research reported in this publication was supported by the National Cancer Institute of the National Institutes of Health under Award Number UM1CA222035. J. Chubak, L.E. Ichikawa, S.A. Merchant, R.A. Ziebell, C.D. Jensen, J.K. Lee, D.A. Corley, N.R. Ghai, B.B. Green, C.S. Skinner, J.E. Schottinger, and T.R. Levin received support from UM1CA222035. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
Abbreviations:
- BMI
body mass index
- CI
Confidence interval
- CRC
colorectal cancer
- FIT
fecal immunochemical test
- gFOBT
guaiac-based fecal occult blood test
- GI
gastrointestinal
- ICD
international classification of diseases
- IBD
inflammatory bowel disease
- KPNC
Kaiser Permanente Northern California
- KPSC
Kaiser Permanente Southern California
- KPWA
Kaiser Permanente Washington
- RD
Risk difference
- USPSTF
United States Preventive Services Task Force
Footnotes
Conflict of interest: Dr. Green is a member of the National Colorectal Cancer Round Table Steering Committee. She receives no compensation for this position, except for travel expenses to national committee meetings.
References
- 1.US Preventive Services Task Force. Screening for Colorectal Cancer: US Preventive Services Task Force Recommendation Statement. JAMA 2021; 325:1965–77. [DOI] [PubMed] [Google Scholar]
- 2.Lin JS, Perdue LA, Henrikson NB, Bean SI, Blasi PR. Screening for Colorectal Cancer: An Evidence Update for the U.S. Preventive Services Task Force. Evidence Synthesis No. 202 ed. Rockville, MD: Agency for Healthcare Research and Quality; 2021. p.45. [PubMed] [Google Scholar]
- 3.Huffstetler AN, Fraiman J, Brownlee S, Stoto MA, Lin KW. An Estimate of Severe Harms Due to Screening Colonoscopy: A Systematic Review. J Am Board Fam Med 2023; 36:493–500. [DOI] [PubMed] [Google Scholar]
- 4.El Halabi J, Burke CA, Hariri E, Rizk M, Macaron C, McMichael J, et al. Frequency of Use and Outcomes of Colonoscopy in Individuals Older Than 75 Years. JAMA Internal Medicine 2023; 183. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Davis AC, Voelkel JL, Remmers CL, Adams JL, McGlynn EA. Comparing Kaiser Permanente Members to the General Population: Implications for Generalizability of Research. Perm J 2023; 27:87–98. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Beaber EF, Kamineni A, Burnett-Hartman AN, Hixon B, Kobrin SC, Li CI, et al. Evaluating and Improving Cancer Screening Process Quality in a Multilevel Context: The PROSPR II Consortium Design and Research Agenda. Cancer Epidemiol Biomarkers Prev 2022; 31:1521–31. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Burnett-Hartman AN, Kamineni A, Corley DA, Singal AG, Halm EA, Rutter CM, et al. Colonoscopy Indication Algorithm Performance Across Diverse Health Care Systems in the PROSPR Consortium. EGEMS (Wash DC) 2019; 7:37. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Deyo RA, Cherkin DC, Ciol MA. Adapting a clinical comorbidity index for use with ICD-9-CM administrative databases. J Clin Epidemiol 1992; 45:613–9. [DOI] [PubMed] [Google Scholar]
- 9.Liang K, Zeger SL. Longitudinal data analysis using generalized linear models. Biometrika 1986; 73:13–22. [Google Scholar]
- 10.Causada-Calo N, Bishay K, Albashir S, Al Mazroui A, Armstrong D. Association Between Age and Complications After Outpatient Colonoscopy. JAMA Netw Open 2020; 3:e208958. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Rutter CM, Johnson E, Miglioretti DL, Mandelson MT, Inadomi J, Buist DS. Adverse events after screening and follow-up colonoscopy. Cancer Causes Control 2012; 23:289–96. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Grossberg LB, Papamichael K, Leffler DA, Sawhney MS, Feuerstein JD. Patients over Age 75 Are at Increased Risk of Emergency Department Visit and Hospitalization Following Colonoscopy. Dig Dis Sci 2020; 65:1964–70. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The data generated in this study are not publicly available due to privacy considerations but may be available upon reasonable request from the corresponding author, pending execution of data sharing agreements (contact corresponding author).