Risk of Incident and Fatal Colorectal Cancer After Young-Onset Adenoma Diagnosis: A National Cohort Study

Abstract

INTRODUCTION: 

Colorectal cancer (CRC) incidence and mortality rates are increasing in adults aged <50 years. Young-onset adenoma (YOA)—adenoma detected in adults younger than 50 years—may signify increased CRC risk, but this association has not been widely studied. Our aim was to compare the risk of incident and fatal CRC in adults aged <50 years with YOA diagnosis compared with those with a normal colonoscopy.

METHODS: 

We conducted a cohort study of US Veterans aged 18–49 years who received colonoscopy between 2005 and 2016. The primary exposure of interest was YOA. Primary outcomes included incident and fatal CRC. We used Kaplan-Meier curves to calculate cumulative incident and fatal CRC risk and Cox models to examine relative CRC risk.

RESULTS: 

The study cohort included 54,284 Veterans aged <50 years exposed to colonoscopy, among whom 13% (n = 7,233) had YOA at start of follow-up. Cumulative 10-year CRC incidence was 0.11% (95% confidence interval [CI]: 0.00%–0.27%) after any adenoma diagnosis, 0.18% (95% CI: 0.02%–0.53%) after advanced YOA diagnosis, 0.10% (95% CI: 0.00%–0.28%) after nonadvanced adenoma diagnosis, and 0.06% (95% CI: 0.02%–0.09%) after normal colonoscopy. Veterans with advanced adenoma had 8-fold greater incident CRC risk than those with normal colonoscopy (hazard ratio: 8.0; 95% CI: 1.8–35.6). Across groups, no differences in fatal CRC risk were observed.

DISCUSSION: 

Young-onset advanced adenoma diagnosis was associated with 8-fold increased incident CRC risk compared with normal colonoscopy. However, cumulative CRC incidence and mortality at 10 years among individuals with either young onset non-advanced or advanced adenoma diagnosis were both relatively low.

INTRODUCTION

Colorectal cancer (CRC) is the third most common cancer and the second leading cause of cancer mortality in the United States, with an estimated 153,020 new cases of CRC and 52,550 CRC deaths in 2023 (¹). Overall, CRC incidence and mortality have decreased over the past 2 decades, but CRC incidence and mortality have increased in young adults aged 18–49 years, with cases often diagnosed at later stages requiring more intense treatment and having poorer prognosis (^2–7).

Most CRCs emerge from adenomas. Adenomas developing in individuals younger than age 50, defined as young-onset adenomas (YOAs), could contribute to early onset CRC risk, though this risk has been incompletely defined. Likewise, strategies for colonoscopy surveillance are not well defined for YOA, and it is unclear whether individuals with YOA might benefit from altered colonoscopy surveillance strategies compared with those without YOA for the purpose of incident and fatal CRC risk reduction. Current surveillance recommendations developed by the US Multi-Society Task Force are based on the findings at baseline colonoscopy, including number, size, and histology of polyps (⁸). Although these guidelines do not recommend differential application of surveillance recommendations based on age, they do note a paucity of data on whether cancer risks differ among individuals with young- vs older-onset adenoma. Evidence on risks for incident and fatal CRC after YOA diagnosis could inform whether current US Multi-Society Task Force guidelines are appropriate for individuals diagnosed with YOA.

Previous studies that have compared CRC risk in young adults vs older adults have faced important limitations, including small sample size, limited generalizability of findings, limited follow-up time, lack of an appropriate comparison group, and focus on advanced adenoma as an outcome rather than CRC (^9–17). Given these limitations, there is a gap in the evidence as to the risk of CRC following YOA diagnosis and whether surveillance guidelines should be tailored based on age. To address this knowledge gap, we examined incident and fatal CRC risk among Veterans aged <50 years with vs without YOA.

METHODS

Study design, setting, and data source

We conducted a cohort study of US Veterans aged 18–49 years receiving care within the Veterans Health Administration (VHA) who received colonoscopy between 2005 and 2016. The VHA is one of the largest integrated health care providers in the United States, providing care to over 6 million individuals annually (¹⁸). Since 1999, all VHA sites have used an integrated electronic health record for documentation of clinical encounters, which can be accessed for research. The Department of Veterans Affairs Corporate Data Warehouse provides access to discrete electronic health record data, including demographic characteristics, administrative claims–based diagnosis and procedure codes, prescriptions, and anthropometric measures (e.g., weight and height), as well as free-text data, including procedure notes and pathology reports, all of which were accessed for this research. We also used data from the VHA Vital Status file to ascertain follow-up time through the date of the last visit, represented as the date and time the last vital record was taken by the health care provider (¹⁹). Linked data from the National Death Index (NDI) were to assess vital status and cause of death. The NDI offers the advantage of capturing cause of death within and outside the VHA. Person-level linkage between VHA data and the NDI cause-specific mortality data was derived through collaboration between VA and Department of Defense partners, with matching based on Social Security number or VA-scrambled Social Security number (²⁰).

Study sample and selection criteria

We included Veterans aged 18–49 years within the VHA with a documented completed colonoscopy between 2005 and 2016. The earliest colonoscopy within this period was considered the baseline colonoscopy. We excluded patients with a history of CRC or inflammatory bowel disease before or at the time of baseline colonoscopy based on the International Classification of Diseases, Ninth Revision diagnosis codes. Veterans whose colonoscopy reports noted an incomplete examination to the cecum or inadequate bowel preparation were also excluded, as were patients with missing data regarding examination extent or bowel prep quality. Veterans who had colonoscopies before 2005 were excluded due to inadequate quality of the colonoscopy and pathology reports.

Exposures, outcomes, and covariates

The primary exposure was YOA identified at baseline colonoscopy. Colonoscopy occurrence was ascertained using a list of Current Procedural Terminology codes derived from a previous study (²¹). Colonoscopy details, including extent of examination and bowel preparation, and YOA details, including location, number, size, and histology, were ascertained via a validated natural language processing–based approach for extracting colonoscopy and pathology findings from free-text reports. This natural language processing-based algorithm was able to identify a normal colonoscopy with 96.3% sensitivity, 97.5% specificity, and a positive predictive value of 97% (²²). Ability to additionally predict location, size, and histology of an adenoma had a positive predictive value of 90% or higher (²³). YOA exposure was defined as any histologically confirmed adenoma at baseline colonoscopy. A normal colonoscopy was defined as a colonoscopy without any documented adenoma or no pathology report with a diagnosis of adenoma or malignancy. Veterans with YOA were categorized as having either (i) advanced adenoma, defined as conventional adenoma ≥10 mm or adenoma with villous histology or high-grade dysplasia of any size, or (ii) nonadvanced adenoma, defined as conventional adenoma <10 mm without any of the histological features of an advanced adenoma.

Primary outcomes included incident and fatal CRC. Incident CRC was ascertained by primary and secondary diagnoses identified through the VA Oncology Domain, which can accurately identify 90% of CRC cases, whereas fatal CRC was ascertained using NDI data in which CRC was listed as the primary cause of death (²⁴). We additionally conducted manual chart review of each suspected CRC case to confirm the date of diagnosis and that the case was identified at follow-up rather than present at baseline colonoscopy. Furthermore, the suspected etiology for each CRC case was classified following the World Endoscopy Organization approach, noting that their categories of “detected lesion, not resected” and “possible missed lesion, prior examination negative but inadequate” are not applicable because we excluded individuals from our cohort with lesions that were not removed at baseline or who had incomplete colonoscopy at baseline (²⁵).

Covariates potentially associated with YOA and CRC risk were selected a priori based on previous literature. These included body mass index, race, ethnicity, aspirin use, diabetes, and smoking status (current, former, or never). We categorized race and ethnicity into 5 different categories: non-Hispanic White, non-Hispanic Black, Asian or Native Hawaiian, American Indian, and Other. Body mass index and diabetes were characterized based on previous identified algorithms (^26,27). Aspirin use was defined by filling of 2 prescriptions or mention of aspirin in free-text notes between 1 year prior to and including the date of baseline colonoscopy (start of follow-up). This methodology was found previously to have a positive predictive value and negative predictive value of 99.2% and 97.5%, respectively (²⁸). Smoking status (current, former, or never) was identified based on VHA Health Factors structured data domain (²⁹).

Statistical analysis

We compared continuous and categorical variables between Veterans with vs without YOA using Wilcoxon rank-sum tests and χ² tests, respectively. We used Kaplan-Meier curves to compare 10-year incident and fatal CRC risk among (i) Veterans with YOA vs those without YOA on baseline colonoscopy, (ii) Veterans with advanced YOA vs without YOA on baseline colonoscopy, and (iii) nonadvanced YOA vs without YOA on baseline colonoscopy. Follow-up started at first (baseline) qualifying colonoscopy and continued until earliest of the following: (i) incident or fatal CRC, non–CRC-related death, or end of the study (December 31, 2016) for the incident CRC analysis or (ii) fatal CRC, non–CRC-related death, or end of the study (December 31, 2016) for the fatal CRC analysis. From these 10-year risk values, we also calculated risk difference and corresponding 95% confidence intervals (CIs). We used Cox models to calculate the incident and fatal CRC risk among adults with advanced YOA or nonadvanced YOA diagnosis, compared with normal colonoscopy. Cox models were used to derive HRs and corresponding 95% CIs. The proportional hazards assumption was tested by examining the correlation between time and scaled Schoenfeld residuals for all exposure variables. The low number of CRC events precluded utilization of covariate-adjusted analyses. We additionally conducted a sensitivity analysis including all individuals who had inadequate bowel prep, incomplete extent of examination, or missing data on bowel prep or extent of examination. We used R version 4.0.2 software to perform statistical analyses. Statistical significance was set at a 2-tailed P value <0.05. The study was approved by the VA San Diego Institutional Review Board.

RESULTS

Cohort characteristics

There were 54,284 Veterans aged 18–49 years who met full study criteria and comprised the analytic cohort (Figure 1), of whom 7,233 (13.3%) had YOA at baseline. Overall, the median follow-up was 4.7 years (Quartile 1–Quartile 3 [Q1–Q3]: 2.1–7.6 years), with the YOA group (median, 3.6 years, Q1–Q3: 1.6–6.4 years) having shorter median follow-up time than the non-YOA group (median 4.9 years, Q1–Q3: 2.2–6.2 years). The median age was 46 years for the adenoma group (Quartile 1–Quartile 3 [Q1–Q3]: 42–48), compared with 43 years for the normal colonoscopy group (Q1–Q3: 35–47). Compared with the normal colonoscopy group, more Veterans in the adenoma group were aged 45–49 years (adenoma vs normal colonoscopy, 50% vs 34%), men (89% vs 80%), obese (50% vs 44%), current smokers (41% vs 33%), and diabetic (11% vs 9%) (Table 1). Baseline polyp characteristics among those with YOA are shown in Supplementary Digital Content (see Supplementary Table 1, https://links.lww.com/AJG/C942).

Figure 1.: Flowchart of the study population with inclusion and exclusion criteria. CRC, colorectal cancer.

Table 1.: Baseline characteristics of individuals with no adenomas vs young-onset adenoma at colonoscopy under age 50 years

YOA exposure and CRC incidence and mortality risk

Among 7,233 Veterans with a baseline diagnosis of any adenoma, 3 (0.08%) developed CRC during the follow-up period. Among 47,051 Veterans with normal colonoscopy, 13 (0.03%) Veterans developed CRC during the follow-up period. Cumulative 10-year CRC incidence was 0.11% (95% CI: 0.00%–0.27%) among adults with YOA at baseline compared with 0.06% (95% CI: 0.02%–0.09%) among adults without baseline YOA. Based on KM curves, the risk of CRC was not significantly different in Veterans with advanced YOA vs no YOA (log-rank P = 0.27) (Figure 2). Cumulative 10-year CRC incidence among Veterans with advanced YOA was 0.18% (95% CI: 0.02%–0.53%) and 0.10% (95% CI: 0.00%–0.28%) among Veterans with nonadvanced adenoma (Table 2; Figure 3). Cumulative 10-year CRC mortality was 0.11% (95% CI: 0.00%–0.32%) among Veterans with advanced YOA and 0.04% (95% CI: 0.01%–0.07%) among Veterans with normal colonoscopy. There were no fatal cases among adults with nonadvanced YOA during the follow-up period. Characteristics of CRC cases found during follow-up are found in Supplementary Digital Content (see Supplementary Table 2, https://links.lww.com/AJG/C942). Among the 3 Veterans with CRC who had baseline adenoma, 2 were categorized as having likely new CRC and 1 as a possibly missed CRC. Among the 13 Veterans with no adenoma at baseline and subsequent CRC, 3 were categorized as likely new and 8 as possibly missed with adequate baseline examination; suspected etiology could not be classified for 2 Veterans.

Figure 2.: Cumulative incidence curves of CRC risk among adults with vs without any adenoma. CRC, colorectal cancer.

Table 2.: Risk for incident and fatal CRC among individuals with any YOA, nonadvanced YOA, and advanced YOA compared with normal colonoscopy under age 50 years

Figure 3.: Cumulative incidence curves of CRC risk among adults with normal colonoscopy, nonadvanced adenoma, or advanced adenoma. CRC, colorectal cancer.

In unadjusted Cox models (Table 2), Veterans with advanced adenoma had 8-fold greater incident CRC risk compared with Veterans with normal colonoscopy (hazard ratio [HR]: 8.0, 95% CI: 1.8–35.6). Having a nonadvanced adenoma was not significantly associated with increased CRC risk compared with those with normal colonoscopy in unadjusted Cox models (HR: 0.8, 95% CI: 0.1–6.1). There was no statistically significant association between advanced adenoma (HR: 6.3, 95% CI: 0.8–50.1) and fatal CRC risk, compared with those with normal colonoscopy, although the CI surrounding this risk estimate was wide. Sensitivity analyses including procedures excluded due to inadequate bowel prep, incomplete extent of examination, or missing information about bowel prep and extent of examination did not produce qualitatively different results (data not shown).

DISCUSSION

Based on this nationwide cohort study of Veterans aged 18–49 years who completed a colonoscopy within the VHA, we report a low overall 10-year cumulative risk of incident or fatal CRC among Veterans with YOA, especially nonadvanced YOA. Although absolute CRC risk was low, Veterans with advanced YOA, but not nonadvanced YOA, have a significantly increased CRC risk compared with Veterans without YOA. Notably, YOA was not associated with fatal CRC risk, although we acknowledge the potential for insufficient power related to the low number of fatal CRC events in this age group. The findings illustrate that although Veterans younger than 50 years with advanced adenoma might be at increased risk for CRC, the low absolute incidence might not necessitate more intense surveillance than recommended for older individuals.

Our study is one of the largest studies to examine CRC risk in adults with vs without YOA. Previous studies limited by small sample size have analyzed metachronous adenoma(s) as an outcome instead of CRC given the rarity of the latter (^9,15,30). A systematic review and meta-analysis conducted by our group included 24 studies and found that the prevalence of YOA was 9%, and the prevalence of metachronous advanced neoplasia after baseline YOA was low at 6% (³¹). Furthermore, studies conducted by Nagpal et al and Hemmasi et al found no statistically significant difference in risk of metachronous adenoma between <50 and ≥50 year-old age groups, although these studies included 128 and 737 total participants, respectively, potentially leading to type II error related to insufficient power (^9,13). Our large study sample enabled us to detect whether a difference in risk existed between those with vs without YOA, filling a critical knowledge gap. That said, the cumulative risk for incident and fatal CRC after YOA diagnosis is low, even among individuals with conventional advanced adenoma (0.18%). For context, based on a study by Lee et al of more than 60,000 patients aged ≥50 years, 10-year CRC risk among patients with advanced adenoma diagnosis was estimated to be 1.2% (³²).

The low absolute CRC risk observed among individuals with young-onset adenoma in our study may suggest that these individuals do not need more aggressive surveillance than what is already recommended for adults aged ≥50 years: 7–10 years of follow-up after diagnosis of 1–2 nonadvanced adenomas <10 mm in size; 3–5 years of follow-up after diagnosis of 3–4 nonadvanced adenomas <10 mm in size; and 3 years of follow-up after diagnosis of an advanced adenoma (⁸). A decision to engage in more aggressive surveillance should also take into account issues of resource utilization, access, and overall health care cost. An editorial in response to the Kim et al 2018 study even argued that surveillance intervals could be lengthened given the low rates of advanced neoplasia in this younger population compared with average-risk adults (³³). Our data may be especially relevant to individuals aged 45–49 years who are newly eligible for screening based on US Preventive Service and US Multi-Society Task Force on CRC recommendations to initiate screening at age 45 years instead of 50 years (³⁴) and have adenomas detected, as our findings would suggest that these individuals do not require more aggressive surveillance than recommended by current polyp surveillance guidelines. This is particularly notable given that 36% of our study population was aged 45–49 years, with half of the group with YOA being aged 45–49 years.

A key strength of our study was utilization of a large national sample, which is markedly higher than previous studies restricted to adults aged <50 years. The use of NDI data, which ascertains cause-specific death data regardless of where adults receive care, also enabled greater confidence in our ascertainment of both incident and fatal CRC cases.

Some limitations should be noted in interpreting this work. Our study population, comprised of US Veterans, was mostly non-Hispanic White individuals and only 19% female, which may raise concerns about generalizability. These concerns can be addressed by future studies assessing risk using non VA data. Despite our large sample size, the outcomes of incident and fatal CRC in our study population were still very rare, leading to small event numbers and possibly limited power to detect differences between groups. Furthermore, the limited number of CRC cases hindered our ability to adjust for key confounders in our models. Thus, important measured confounders, such as age, sex, and race/ethnicity, and unmeasured covariates, such as diet, family history, alcohol use, baseline procedure indication, and environmental/military exposures, could not be accounted for in our risk estimates. The available sample size also precluded subgroup analyses stratified by characteristics such as age group. Our natural language processing algorithm to ascertain colonoscopy-related information relied on high-quality reports that may have not been as sensitive for collection of data on bowel prep and extent of examination, leading to a high number of individuals with missing information. To account for this, we conducted a sensitivity analysis including all excluded individuals with missing bowel prep, inadequate bowel prep, or incomplete extent of examination, finding results that were not qualitatively different from the primary analyses. Our study did not examine exposure to colonoscopy after baseline, which might have provided greater context about follow-up of baseline findings and natural history of CRC in younger adults. Future studies should examine how polyp surveillance might impact CRC incidence and mortality, particularly among adults found to have an adenoma at their initial colonoscopy.

Among adults younger than 50 years with YOA diagnosis, the risk for incident and fatal CRC is low, even among individuals with baseline advanced adenoma. Compared with individuals with normal colonoscopy, risk for incident CRC was similar for patients with nonadvanced YOA and higher for patients with advanced YOA. Taken together, these data suggest that patients with YOA are unlikely to account for a substantial proportion of early-onset CRC diagnoses and that individuals with YOA may not need to have surveillance that is more aggressive than currently recommended for older individuals with adenomas.

CONFLICTS OF INTEREST

Guarantor of the article: Samir Gupta, MD, MSCS.

Specific author contributions: Y.C., J.D., N.E., L.L., R.B., and S.G.: concept and design. Y.C., J.D., L.L., R.B., S.M., S.S., and S.G.: analysis and interpretation of data. Y.C., J.D., and S.G.: drafting of the manuscript. Y.C., J.D., N.E., L.L., C.J., A.E., R.B., S.M., S.S., and S.G.: critical revision of the manuscript for important intellectual content. Y.C., J.D., L.L., R.B., and S.G.: statistical analysis. S.M.: graphical abstract. J.D. and S.G.: obtained funding.

Financial support: This research was supported by Grant No. 5F32CA239360-02 (PI: J.D.) from the National Cancer Institute/National Institutes of Health; Grant No. 1K99CA267181-01A1 (PI: J.D.) from the National Cancer Institute/National Institutes of Health; Grant No. 5I01HX001574-05 (PI: S.G.) from VA Health Services Research and Development; Grant No. 5R37CA222866-02 (PI: S.G.) from the National Cancer Institute/National Institutes of Health; and Grant No. T32DK007202 from the National Institute of Diabetes and Digestive and Kidney Diseases/National Institutes of Health (PI: S.G.).

Potential competing interests: None to report.