Endoscopic Screening and Risk of Colorectal Cancer according to Type 2 Diabetes Status

Colorectal cancer is the second-leading cause of cancer death in the combination of men and women in the United States (1). Endoscopic screening has been shown to reduce colorectal cancer incidence and mortality by identifying and removing precancerous polyps and early-stage cancers (2–7). Despite an overall increase in the uptake of endoscopic screening over the years (8), approximately 40% screening-eligible adults in the United States do not comply, and there is also substantial disparity in the prevalence of screening (9, 10). A concerning increase in the incidence of early-onset colorectal cancer (diagnosed at <50 years of age; refs. 1, 11, 12) prompted the U.S. Preventive Services Task Force (USPSTF) to lower the recommended age to initiate colorectal cancer screening from 50 to 45 years in 2021 (13). However, there is still a debate about whether such a recommendation of lowering the starting age for all individuals regardless of risk would induce high financial burdens and a more unbalanced screening resource allocation (14, 15). Numerous studies have developed screening recommendations based on colorectal cancer risk stratification by prediction models (16–21), but it remains unclear whether screening is similarly effective across the risk spectra. Profiling the practical benefit of endoscopic screening according to more colorectal cancer–associated characteristics could inform a better personalized recommendation for colorectal cancer screening.

In the past decades, the United States has seen an increase in the prevalence of diabetes mellitus, growing from 9.5% in 2000 to 13.0% in 2018 among adults (22). Type 2 diabetes (T2D) is most often diagnosed in elderly individuals, but notably, the incidence of early-onset T2D (diagnosed at <40 years of age) is rising in the United States and worldwide (23, 24). From 2013 to 2017, the U.S. prevalence of T2D in individuals ages <40 years increased from 3.5% to 4.3% (24). The presence of T2D has been associated with an increased colorectal cancer risk (25, 26). Cancer was recently reported to have overtaken vascular disease as the leading cause of death in T2D individuals (27, 28). However, the current recommendation for colorectal cancer screening was based only on age and family history, whether the T2D population benefits more from endoscopic screening than the general non-T2D population and how the age to start colorectal cancer screening may be tailored for T2D patients remain unknown. Addressing these questions has significant clinical implications for improved colorectal cancer prevention.

Therefore, in the current study, we prospectively assessed the relative and absolute risk of colorectal cancer incidence and mortality associated with endoscopic screening according to individuals’ T2D status within 3 large cohorts in the United States. We also examined the age-specific colorectal cancer cumulative incidence and mortality and identified the age when the threshold colorectal cancer risk at age 45 and 50 years in non-T2D individuals was attained among individuals with T2D.

Data of three ongoing prospective cohorts were used this study, including the Health Professionals Follow-up Study (HPFS), the Nurses’ Health Study (NHS), and the NHS II. The HPFS enrolled 51,529 male health professionals ages 40 to 75 years in 1986. The NHS and NHS II enrolled 121,701 female nurses ages 30 to 55 years in 1976 and 116,429 female nurses ages 25 to 42 years in 1989, respectively (29–31). Detailed information of lifestyle, medical history, and other characteristics was collected by questionnaire in all 3 cohorts at baseline and biennially thereafter with a follow-up rate of more than 90%. In this study, the year of 1988 for the HPFS and NHS and 1991 for the NHS II were used as baseline, when we first collected detailed information on endoscopy. We conducted a baseline exclusion of the participants who had inflammatory bowel disease, cancer (except nonmelanoma skin cancer), missing information on endoscopy, or missing information on major lifestyle factors. As a result, 42,855 men from the HPFS, 76,781 women from the NHS, and 89,536 women from the NHS II were included in the analysis. Written informed consent was obtained from the participants. This study has been performed in accordance with the Declaration of Helsinki. The study protocol was approved by the institutional review board of the Brigham and Women's Hospital and Harvard T.H. Chan School of Public Health, and those of participating registries as required.

Beginning in 1988 and continuing through 2002 in the HPFS and NHS, 1991 through 2001 in the NHS II, participants were asked biennially whether they had undergone lower gastrointestinal endoscopy and, if so, the reason for the endoscopy. In 2004 in the HPFS and NHS and 2003 in the NHS II, whether the previously reported endoscopies were sigmoidoscopies or colonoscopies was additionally inquired. After that, we separately recorded sigmoidoscopy and colonoscopy in every cycle. We defined an endoscopic screening as any lower gastrointestinal tract endoscopy for asymptomatic or routine screening or because of a family history of colorectal cancer (5). In random samples of participants who reported having had negative endoscopy (n = 140 in the HPFS and 114 in the NHS), we observed a high concordance rate (97%–100%) between self-report and endoscopic records (32–34), indicating a good accuracy of the self-reported data.

For participants who reported a diagnosis of T2D in the cohort questionnaire, we mailed them a supplementary questionnaire to inquire about symptoms, diagnostic tests, and hypoglycemic therapy. We used the National Diabetes Data Group criteria to confirm the T2D cases diagnosed before 1998 (35). For T2D cases after 1998, the diagnosis was confirmed by fasting plasma glucose of ≥7.0 mmol/L on two separate tests according to the American Diabetes Association criteria. For T2D cases after January 2010, we further considered HbA1c of ≥6.5% in the diagnosis criteria (36). The return date of the follow-up questionnaire in which participants first reported T2D diagnosis was used as the diagnosis date if the diagnosis date was unknown. The validity of using our supplementary questionnaire to adjudicate T2D diagnosis has been verified in our validation studies (97% in the HPFS and 98% in the NHS; refs. 37, 38).

When colorectal cancer diagnosis was reported on biennial questionnaires, we examined medical records and pathology reports to confirm the diagnosis under written permission from participants. We also identified unreported fatal colorectal cancer cases through family members or the postal system, or sought from the National Death Index, tumor registries, and death certificates. All information was reviewed by the study physicians blinded to exposure information to confirm colorectal cancer diagnosis, including anatomic location and stage. Colorectal cancers were classified into proximal colon cancers occurring in the cecum, ascending colon, hepatic flexure, and transverse colon, distal colon cancers occurring in the splenic flexure, descending colon, and sigmoid colon, and rectal cancers occurring in the rectosigmoid junction and rectum. Colorectal cancer stage was defined in accordance with the American Joint Committee on Cancer (AJCC) tumor–node–metastasis (TNM) cancer staging system (39).

We obtained information on major risk factors for colorectal cancer at baseline and biennial follow-up questionnaires, including ethnicity, cigarette smoking, body mass index (BMI), colorectal cancer history in first-degree relatives, physical activity, uses of aspirin and multivitamin, menopausal status, and hormone use (women only). Physical activity was evaluated by total hours per week for moderate-to-vigorous intensity activity (including brisk walking) requiring ≥3 metabolic equivalents per hour. Dietary information was collected every 4 years using validated food frequency questionnaires (40). We used the six major dietary recommendations in the World Cancer Research Fund/American Institute for Cancer Research (WCRF/AICR) Third Expert Report released in 2018 (41) to evaluate diet quality (range, 0–6), including a daily consumption of red meat <0.5 servings, processed meat <0.2 servings, dietary fiber ≥30 g, dairy products ≥3 servings, whole grains ≥48 g or accounting for ≥50% total grains, and calcium intake ≥1,000 mg.

Person-time was counted from return of the baseline questionnaire (1988 for HPFS and NHS and 1991 for NHS II) to colorectal cancer diagnosis date (for colorectal cancer incidence analysis only), death, loss to follow-up, or end of the follow-up period (HPFS: January 31, 2016; NHS: June 30, 2016; NHS II: June 30, 2017), whichever came first. Statuses of endoscopic screening and T2D were defined in a time-varying form, i.e., participants were considered non-screening and non-T2D until the first time when they reported screening (or a diagnosis of T2D), and were considered screening and non-T2D (or non-screening and T2D), until their change to the next status of screening and T2D, and remained in this status thereafter for the remainder of follow-up. To avoid counting in the screening group the colorectal cancer cases that were detected at screening but had likely occurred before screening, we stopped updating endoscopic screening information one questionnaire cycle earlier before colorectal cancer diagnosis (only for colorectal cancer incidence analysis). Because of the focus on screening endoscopy in this study, in mortality analysis, we stopped updating endoscopy information once a participant was diagnosed with colorectal cancer.

Age-, questionnaire cycle-, and cohort-stratified and time-varying Cox proportional hazards regression models were used to estimate the multivariable hazard ratios (HR) and 95% confidence intervals (CI) for the association between endoscopic screening and colorectal cancer incidence and mortality according to the status of T2D. To examine the absolute benefit of endoscopic screening according to T2D status, we calculated the multivariable-adjusted colorectal cancer cumulative incidence and mortality in the screening and non-screening group separately according to T2D status using Cox regression models with age as the time scale. Given that colorectal cancer screening is currently recommended until age 75 years, we calculated the cumulative risks by age 80 years. In each of non-T2D and T2D groups, absolute risk reduction (ARR) associated with endoscopic screening was calculated by subtracting the cumulative risk in the screening group from that in the non-screening group.

Multiplicative interaction in the HRs by T2D status was tested through including in the model the main effects of endoscopic screening and T2D status as well as their product term. We considered P value of the product term as the P for multiplicative interaction. We assessed the additive interaction of the ARRs by T2D status through regressing the multivariable-adjusted cumulative risk on endoscopic screening and T2D status as well as their product term, whose P value was considered as the P for additive interaction. We also examined difference in the interactions by T2D between women and men, between healthy weight (BMI <25 kg/m²) and overweight or obese individuals (BMI ≥25 kg/m²), and between those with and without a family history of colorectal cancer (three-way interaction). The three-way interaction in the ARR by T2D and sex was assessed by regressing the multivariable-adjusted cumulative risk on endoscopic screening, T2D status, sex, three product terms of each two of the three, and the product term of the three, and P value of the three-term product was considered as the P for three-way interaction. The same method was used in assessing the three-way interactions of T2D with body weight and family history of colorectal cancer.

We then examined the risk-adapted starting age of colorectal cancer screening for individuals with T2D. The risk-adapted screening-starting age was defined as the age at which T2D individuals are expected to attain the threshold colorectal cancer risk level that was reached by the general non-T2D individuals at age 45 years, the generally recommended age to start colorectal cancer screening (42, 43). Age-specific cumulative colorectal cancer incidences were estimated using multivariable Cox proportional regression models in the T2D and non-T2D group separately. Baseline age for the non-T2D group and T2D diagnosis age for the T2D group were used as the age scale. Due to the natural history of colorectal cancer and extended protection of endoscopic screening (3, 44), both the 10- and 20-year cumulative colorectal cancer incidence were used as the risk scale. In addition, we used the 10- and 20-year cumulative colorectal cancer mortality as the risk scale in sensitivity analysis.

We conducted the analyses using SAS version 9.4 (SAS Institute) and considered two-sided P values of <0.05 as statistically significant.

Because of the sensitive nature of the data collected for this study, requests to access the data set from qualified researchers trained in human subject confidentiality protocols may be sent to Brigham and Women's/Harvard Cohorts at https://sites.google.com/channing.harvard.edu/cohortdocs/getting-started/collaborations-consortia?authuser=2.

Mean (SD) age at baseline was 47 (11) years. In the three cohorts of 209,172 participants with a median of 26 years (up to 28 years and 5,040,736 person-years) of follow-up, we documented 3,457 incident colorectal cancer cases and 1,129 colorectal cancer deaths, among which 369 cases (11%) and 134 deaths (12%) occurred in participants with T2D, and 2,921 cases (84%) and 838 deaths (74%) occurred at or before age 80 years. By subsite, 1,313 cases (38%) and 423 deaths (37%) were of proximal colon cancer, 853 cases (25%) and 257 deaths (23%) of distal colon cancer, and 725 cases (21%) and 252 deaths (22%) of rectal cancer. By the end of the follow-up, 13% of the participants had been diagnosed with T2D, and 63% had reported a history of endoscopic screening. Based on the person-years over follow-up (Table 1), the prevalence of T2D was 6% and of endoscopic screening was 32%. In the non-T2D and T2D groups, the prevalence of endoscopic screening was 31% and 45%, respectively. However, we found a similar proportion of advanced-stage colorectal cancer (TNM 3 and 4) in non-T2D (54%) and T2D groups (56%).

Overall, we observed an increased cumulative incidence of colorectal cancer in the T2D group (3.33%) compared with non-T2D group (2.68%; P < 0.001; Supplementary Table S1). In both groups of non-T2D and T2D, individuals with endoscopic screening had a significantly lower cumulative incidence of colorectal cancer than the non-screening group (screening vs. non-screening = 1.83 vs. 3.56% in non-T2D and 2.07 vs. 4.43% in T2D; Fig. 1). When we look at the reduction in the absolute risk, we found the endoscopic screening-associated ARR for colorectal cancer incidence was significantly higher in the T2D group (2.36%) than in the non-T2D group (1.73%; P-additive interaction = 0.01), which amounted to 42 T2D individuals and 58 non-T2D individuals, respectively, needed to screen (number needed to screen) to prevent one incident colorectal cancer case. We did not find a significant interaction of endoscopic screening-associated HR for colorectal cancer incidence by T2D status (P-multiplicative interaction = 0.57; Supplementary Tables S2 and S3).

For colorectal cancer mortality (Fig. 1), individuals in the screening group had a significantly lower cumulative colorectal cancer mortality than the non-screening group in both non-T2D and T2D individuals (non-T2D: 0.40 vs. 1.16%; T2D: 0.56 vs. 1.74%). Consistent with the finding for colorectal cancer incidence, the screening-associated ARR for colorectal cancer mortality was also higher in T2D (1.17%) than in non-T2D individuals (0.76%; P-additive interaction = 0.003). Different from that for colorectal cancer incidence, we found a stronger protective HR associated with endoscopic screening for colorectal cancer mortality in T2D (0.37; 95% CI, 0.32–0.41) than in non-T2D individuals (0.47; 95% CI, 0.43–0.51; P-multiplicative interaction = 0.01), and the interaction was stronger in men than in women (Supplementary Tables S2 and S3).

When we compared the screening-associated ARRs between the non-T2D and T2D groups by colorectal cancer subsite (Fig. 2), we found the increased ARR in T2D compared with non-T2D individuals was stronger for proximal colon cancer (P interaction = 0.001) and distal colon cancer (P interaction = 0.01) than for rectal cancer (P-additive interaction = 0.95). Also, as shown in Table 2, we found the increased benefit of endoscopic screening in T2D individuals was higher in men (P interaction by T2D <0.001) than in women (P interaction by T2D = 0.02; P for three-way interaction by T2D and sex = 0.01). However, the increased benefit in T2D individuals was similar across those with healthy body weight and overweight or obesity (P for three-way interaction by T2D and BMI = 0.81), and those with and without a family history of colorectal cancer (P for three-way interaction by T2D and sex = 0.89).

As shown in Fig. 3, the 10-year cumulative colorectal cancer incidence increased substantially with age, and the T2D group had a higher incidence than the non-T2D group at each age point. At the benchmark age of 45 years, when average-risk adults are generally recommended to start regular screening for colorectal cancer, the non-T2D individuals attained a 10-year cumulative colorectal cancer incidence of 0.35%. In contrast, individuals with T2D might reach this threshold risk level 9 years earlier, at age 36 years. When we used the age of 50 years as the benchmark, the 10-year cumulative colorectal cancer incidence attained by the non-T2D individuals was 0.51%, and the T2D individuals reached this threshold level also 9 years earlier, at age 41 years. When we used 20-year cumulative colorectal cancer incidence and 10- and 20-year cumulative colorectal cancer mortality as the risk scale, we found that T2D individuals might start colorectal cancer screening 5, 11, and 7 years earlier than the recommended age of 45 years, respectively (Supplementary Figs. S1–S3).

In three large prospective cohorts, we found that, although endoscopic screening was associated with a similar relative risk of lower colorectal cancer incidence between T2D and non-T2D individuals, the absolute risk of colorectal cancer incidence and mortality reduced by endoscopic screening was substantially higher in T2D than non-T2D individuals. This enhanced absolute benefit of endoscopic screening associated with T2D was higher for cancers in the proximal and distal colon than the rectum, and higher in men than women. Also, we observed that individuals at the age of 45 years without T2D attained a 10-year cumulative colorectal cancer risk of 0.35%, whereas individuals with T2D at the age 36 years might reach this risk level, suggesting that the T2D individuals may start colorectal cancer screening up to 9 years earlier than the generally recommended 45 years old. These findings provide evidence for the development of tailored colorectal cancer screening recommendations based on T2D status.

Benefit of endoscopic screening in the prevention of colorectal cancer and related deaths has been widely reported (2–7). We previously reported that individuals with a higher colorectal cancer risk profile, defined by family history of colorectal cancer, height, BMI, smoking, alcohol intake, physical activity, diet, and aspirin use, might benefit more from endoscopic screening than those with a lower risk profile (43). However, given the growing prevalence of T2D and increased colorectal cancer risk in the T2D population, no prior study has investigated whether the benefit of endoscopic screening differs between individuals with and without T2D. We found evidence supporting a greater benefit of endoscopic screening in T2D individuals. Also, we found that individuals with T2D may benefit from an earlier initiation (up to 9 years) of regular colorectal cancer screening than the generally recommended age of 45 years. It is largely in line with a recent study of the Swedish registries (45), which reported that T2D individuals might start colorectal cancer screening 5 to 17 years earlier than the generally recommended screening-starting age. Given the rising incidence of early-onset T2D and early-onset colorectal cancer (23, 24), our findings add to the growing evidence supporting early colorectal cancer screening in the young-onset diabetic population.

Endoscopic screening and T2D influence colorectal cancer risk through apparently independent pathways, i.e., endoscopic screening protects against incident colorectal cancer through detection and removal of colorectal precancerous lesions, whereas insulin resistance and hyperinsulinemia were suggested as the mechanism linking T2D and increased colorectal cancer risk (25). Therefore, it is not surprising that we observed relatively consistent hazard ratios associated with endoscopic screening for colorectal cancer risk between T2D and non-T2D individuals. However, given the effect of T2D on increasing colorectal cancer risk, individuals with T2D are at higher risk of developing colorectal precancerous lesions and early colorectal cancer (46–48). It is thus understandable that endoscopic screening confers a higher reduction in the absolute risk of colorectal cancer incidence and mortality in T2D than in non-T2D individuals.

Our findings have important clinical and public health implications. First, although relative risk is important to determine the effect of screening, ARR may be more important for clinical decision-making owing to its indication for priorities in health-care service (49). Given that the current recommendations for colorectal cancer screening are the same for all except for those with a family history of colorectal cancer (1), our findings provide empirical evidence for setting screening priorities according to T2D status and add to the growing data supporting the development of risk-based screening strategies to improve health-care delivery for better colorectal cancer prevention. Second, we found that the difference in the screening-associated ARR between T2D and non-T2D individuals was more pronounced for colon than rectal cancer and more pronounced in men than women. Given the diminished benefit of endoscopy in preventing tumors in the colon than the rectum (3), our findings suggest the particular importance of high-quality endoscopic screening among individuals with T2D. For the observed difference between men and women, potential reasons may include the higher prevalence of T2D (22), higher colorectal cancer incidence, and higher prevalence of colorectal cancer risk factors in men than women (1). However, given the small number of cases in some of the strata, more studies are needed to replicate these subgroup findings. Third, the recently updated recommendations by the USPSTF and American Cancer Society (13, 50) of lowering colorectal cancer screening starting age from 50 to 45 years for all individuals regardless of risk may induce high financial burdens and a more unbalanced screening resource allocation (14, 15). Our findings provide evidence supporting a tailored colorectal cancer screening recommendation that would more precisely target the population who is more likely to benefit from an earlier screening.

Our study has several strengths, including the large sample size, long-term follow-up, repeated assessments of lifestyle and endoscopy, and medical record-confirmed disease diagnoses. Several limitations should also be noted. First, the information of lifestyle factors and endoscopic screening was self-reported and thus subject to measurement error. However, the accuracy of these self-reported data in our cohorts has been well documented (32–34, 51, 52). Second, our study participants are health professionals and predominantly Whites, thereby limiting the generalizability of our conclusion. More studies, particularly clinical trial studies, are needed to confirm our findings about the added benefit of endoscopic screening in T2D patients. Third, in the calculation of the risk-adapted starting age of screening, we used the healthy non-T2D participants to generate the threshold colorectal cancer risk. Because our cohort participants have a better health profile compared with the general population (53), we may have underestimated the threshold colorectal cancer risk. Also, because of a higher frequency of hospital visit, T2D individuals may be more likely to undergo screening and have colorectal cancer diagnosed at a younger age than those without T2D. Although we found a similar proportion of advanced-stage colorectal cancer (TNM 3 and 4) in T2D (54%) and non-T2D individuals (56%), we adopted a two-year lag analysis to further avoid the issue of earlier colorectal cancer detection in T2D individuals.

In conclusion, the absolute benefits of endoscopic screening in preventing colorectal cancer and related death are higher in individuals with T2D than those without T2D. Individuals with T2D may consider starting colorectal cancer screening up to 9 years earlier than the recommended age of 45 years. Our findings provide evidence supporting a T2D-tailored recommendation for colorectal cancer screening.

This study was started while Dr. Wu was still an employee at the Harvard T. H. Chan School of Public Health. Dr. Kana Wu is now an employee of and holds stocks in Vertex Pharmaceuticals. This study was not funded by this commercial entity. K. Ng reports grants from the NCI during the conduct of the study; grants from Colorectal Cancer Alliance, nonfinancial support from Pharmavite, grants from Evergrande Group, Janssen, Revolution Medicines, personal fees from Bayer, Seattle Genetics, Array Biopharma, BiomX, Bicara Therapeutics, GSK, Pfizer, X-Biotix Therapeutics, and Redesign Health outside the submitted work. M. Wang reports grants from NIH during the conduct of the study. S. Ogino reports grants from NIH during the conduct of the study. A.T. Chan reports grants from Zoe Ltd, Freenome, personal fees from Bayer Pharma AG, Boehringer Ingelheim, and grants and personal fees from Pfizer Inc. outside the submitted work. No disclosures were reported by the other authors.

K. Wang: Conceptualization, data curation, software, formal analysis, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. W. Ma: Methodology, writing–review and editing. Y. Hu: Methodology, writing–review and editing. M.D. Knudsen: Methodology, writing–review and editing. L.H. Nguyen: Methodology, writing–review and editing. K. Wu: Methodology, writing–review and editing. K. Ng: Methodology, writing–review and editing. M. Wang: Methodology, writing–review and editing. S. Ogino: Methodology, writing–review and editing. Q. Sun: Methodology, writing–review and editing. E.L. Giovannucci: Methodology, writing–review and editing. A.T. Chan: Funding acquisition, methodology, writing–review and editing. M. Song: Conceptualization, resources, data curation, software, supervision, funding acquisition, investigation, methodology, project administration, writing–review and editing.

This work was supported by the NIH grants (P01 CA87969 to M.J. Stampfer; U01 CA186107 to M.J. Stampfer; P01 CA55075 to W.C. Willett; UM1 CA167552 to W.C. Willett; U01 CA167552 to L.A. Mucci and W.C. Willett; K24 DK098311, R01 CA137178, R01 CA202704, and R01 CA176726 to A.T. Chan.; K99 CA215314, R00 CA215314, U01 CA261961, and R01 CA263776 to M. Song; R35 CA197735 and R01 CA151993 to S. Ogino); by the American Cancer Society Mentored Research Scholar Grant (MRSG-17-220-01-NEC to M. Song). A.T. Chan is a Stuart and Suzanne Steele MGH Research Scholar. The funders had no role in the design and conduct of the study; collection, management, analysis and interpretation of the data; preparation, review, or approval of the manuscript; or decision to submit the manuscript for publication. The authors thank the participants and staff of the HPFS, the NHS, and the NHS II for their continued contributions, as well as the following state cancer registries for their help: Alabama, Arizona, Arkansas, California, Colorado, Connecticut, Delaware, Florida, Georgia, Idaho, Illinois, Indiana, Iowa, Kentucky, Louisiana, Maine, Maryland, Massachusetts, Michigan, Nebraska, New Hampshire, New Jersey, New York, North Carolina, North Dakota, Ohio, Oklahoma, Oregon, Pennsylvania, Rhode Island, South Carolina, Tennessee, Texas, Virginia, Washington, and Wyoming. The authors assume full responsibility for analyses and interpretation of these data.

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.