The relationship between diabetes and the risk of colon cancer (CC), two highly prevalent and major health problems worldwide, is well recognized. Most previous epidemiologic studies have provided evidence that individuals with diabetes have an increased risk of CC compared with their nondiabetic counterparts (1–4). An updated 2011 meta-analysis suggested that diabetes was an independent risk factor for colon and rectal cancer, after adjusting major risk factors including obesity, smoking and physical activity, with a corresponding summary relative risk (RR) of 1.38 and 1.20, respectively (5). The so-called “hyperinsulinemia hypothesis”, which suggests that elevated levels of insulin and insulin-like growth factor-1 (IGF-1) increase the risk of CC by promoting the growth of colon cells and acting as a cell mitogen, provides the underlying mechanism for this connection (6).

Although extensive research has been performed on this topic, several features of the association between diabetes and the risk of CC remain unclear. For instance, it is not known whether diabetes is differentially associated with the risks of right-sided colon cancer (RCC) and left-sided colon cancer (LCC). A growing body of evidence suggests that LCC and RCC should be considered two distinct clinical and biological tumor entities. The left side of the colon originates from the midgut, whereas the right side originates from the hindgut. Differences also exist that correlate to physical function, artery supply, histology and biochemical features (7, 8). Subsequent research has shown that there are distinct differences in epidemiology (9), pathogenesis, genetic landscape, molecular pathways (10), and the clinical outcome (11, 12) between cancers at these two anatomical sites.

Understanding whether diabetes is differentially associated with the risks of developing RCC and LCC may have important clinical and public health implications. Screening for colorectal cancer (CRC) in average-risk or high-risk patients using flexible sigmoidoscopy (FS) or colonoscopy is now common in many countries. However, FS screens only the distal colon (10). In 2005, Limburg and colleagues first reported that diabetes was significantly associated with the risk of RCC, but not LCC, and suggested that CRC screening methods should include evaluation of the right colon for patients with diabetes to improve the effectiveness of CRC prevention (13). However, findings from other previous studies have been inconsistent (1, 14), and there has been no systematic comparison of subsite differences between diabetes and CC risk.

Given the rising prevalence of diabetes as a global health problem and the substantial clinical implications that any important subsite difference in the association between diabetes and CC risk would have, we undertook a meta-analysis of all available prospective cohort studies that reported the site-specific effects of diabetes on the subsequent risk of CC.

This study was conducted according to the Meta-analysis Of Observational Studies in Epidemiology (MOOSE) guidelines (15).

In this pooled analysis of prospective cohort studies, with data from 1,642,823 individuals and 17,624 CC events, we examined the site-specific association between diabetes and the risk of CC. Our findings support a role for diabetes in the etiology of CC, including both RCC and LCC. After adjusting for major risk factors, individuals with diabetes showed a significant 13% increased risk of RCC compared with LCC.

The main strength of this pooled analysis is that by including only prospective cohort studies, we were able to perform an objective meta-analysis based on a large population size over a long follow-up duration. The ten large prospective cohort studies included better resemble clinical practice, are of good quality and lack obvious selection bias compared with case-control studies and retrospective cohorts. Our findings were robust with no heterogeneity across studies and no obvious evidence of publications bias. Moreover, our results were consistent in a range of subgroup analyses.

Diabetes and CC share similar risk factors, including obesity, physical activity and the Western diet. Thus, our results could be confounded by these risk factors. However, a previous meta-analysis found that BMI and meat consumption were associated with a more pronounced risk of LCC, while physical activity and other dietary risk factors did not differ between LCC and RCC (39–41). In our analysis, nine out of ten studies adjusted for major risk factors (≥ 5, including obesity, smoking and physical activity) of CC in addition to age. Importantly, there was no evidence of a significant difference between age-adjusted RR/RRRs and multi-adjusted RR/RRRs. Thus, it is unlikely that the results were influenced by confounders. In addition, we directly compared the risk of RCC and LCC from within the same study, thereby reducing the role of extraneous, between-study known and unknown confounding factors.

Nevertheless, several limitations merit comment. First, there are racial/ethnic differences in the incidence of CRC. According to the 2014 Surveillance, Epidemiology, and End Results (SEER) Cancer Statistics, CRC incidence rates were highest among blacks, followed by American Indians/Alaska Natives, non-Hispanic whites, Hispanics and Asian/Pacific Islanders (42). In our study, except for the multiethnic cohort, the patients included were predominantly white from the other nine cohorts. Further studies are warranted to address whether this increased risk of RCC compared with that of LCC remains in other races/ethnicities. Second, diabetes status was based on self-report, which could have resulted in some misclassification of diabetics as nondiabetics. Although previous studies have shown that self-reported diabetes is an accurate proxy compared with medical records (43, 44), such misclassification still could underestimate the true relationship between diabetes and CC. Third, the studies included in this analysis did not distinguish between type 1 and type 2 diabetes. As type 1 diabetes accounts for 5-10% of all cases of diabetes and may not be related to CC (45), associations were likely to be underestimated in this study. Nevertheless, the incidence of type 1 diabetes increases with age, peaking at around 10–14 years (46). The greatest observed increases in the incidence of type 1 diabetes are among children younger than 15 years, particularly those younger than 5 years (47). In our study, all patients enrolled were ≥ 30 years old, thus they were much less likely to have type 1 diabetes than type 2 diabetes. Fourth, not all of the CC cases could be classified by location, which may have influenced the real RRs or RRRs. In our analysis, we only included studies which classified at least 80% of CRC cases or CC cases by subsite. Fifth, most of the studies involved in this analysis did not report on, or adjust for antidiabetic drug use, which could influence the true associations between diabetes and CC (48). However, the relationship between the use of anti-diabetic drugs and the incidence of colon cancer is still unclear (49–52). Previous meta-analyses suggested that metformin potentially reduces the risk of colon cancer in patients with type 2 diabetes (53–55), while insulin use was associated with an increased risk (50, 51). Some preclinical studies also showed that the SGLT2 inhibitors might attenuate colon cancer cells growth (56–59).

The hyperinsulinemia theory, which implies that elevated insulin and free IGF-1 levels are the two key components that promote the growth of colon tumors (60), may relate diabetes to colon cancer. Insulin and IGF-1 receptors are widely distributed in normal colonic epithelium and colon cancer tissue (61, 62). Insulin is an important growth factor for colonic mucosal cells (63). Preclinical studies have shown that insulin promotes not only the growth and survival of colon cancer cells but also the biosynthesis of IGF-I (64), while IGF-1 inhibits the apoptosis of colon cancer cells (65). There is some potential biologic evidence to explain why diabetes may have different associations with the risks of RCC and LCC. Leptin, which is regulated by insulin (66), has been shown to increase colonic cell proliferation and stimulate DNA synthesis in the proximal colon, but not in the distal colon (67). In an IGF-1-deficient rodent model, reduction of IGF-1 altered the location of the colonic tumor. Significant inhibition of colon tumor multiplicity in the proximal colon was observed compared to that in the distal colon (68). The distinct sensitivity to leptin or IGF-1 between proximal and distal colon cells implies differences in genetic susceptibilities to carcinogens through different molecular mechanisms.

In 2014, the worldwide prevalence of type 2 diabetes was approximately 422 million, accounting for 8.4% of adults, and this rate is projected to increase rapidly. In 2019, diabetes was the direct cause of 1.5 million deaths (69). The present study provides the most comprehensive assessment of possible site differences between diabetes and CC risk. Our findings may have important clinical implications for customized CRC screening programs in diabetic patients. CRC screening tests using FS or colonoscopy are now common in many countries and have been proven effective to reduce CRC incidence and mortality. Although FS is more widely available, quicker, more convenient and less expensive than colonoscopy, it allows visualization of only the distal part of the colon, while colonoscopy allows visualization of the entire colon (10, 70). Hence, individuals who are at high risk for proximal lesions, such as diabetic patients, are more suited for colonoscopy; FS may be of less value. Our findings reinforce the importance of stratifying for high-risk populations, such as patients with diabetes to improve the effectiveness of CRC screening and prevention. Close colonoscopic surveillance in diabetic patients with careful examination of the right colon is warranted.

To conclude, the results from this meta-analysis suggest that diabetes is associated with increased risks for both RCC and LCC and that the association between diabetes and the risk of RCC is significantly higher than the risk of LCC after adjusting for major risk factors. Our findings add to the increasing evidence for the distinct clinical and biological patterns between RCC and LCC and provide further support for public health efforts aiming to establish a more tailored approach to CRC screening strategies in diabetic patients.

The original contributions presented in the study are included in the article/ Supplementary Material . Further inquiries can be directed to the corresponding author.

WX and BW contributed to the study concept and design, data analysis, interpretation of the data and editing the manuscript. WX and JH contributed to the data acquisition, statistical analysis and editing the manuscript. WX, JH, CZ and LD contributed to the statistical analysis. WX, JH and XW contributed to the interpretation of the data. All authors commented on drafts of the paper and have approved the final draft of the manuscript for submission.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.