Preexisting Type 2 Diabetes and Survival among Patients with Colorectal Cancer

Type 2 diabetes (henceforth referred to as “diabetes”) is known to increase risk of developing colorectal cancer, the second leading cause of cancer-related deaths in the United States (1). In a meta-analysis of 15 studies, individuals with diabetes had a RR for colorectal cancer of 1.30 [95% confidence interval (CI), 1.20–1.40], compared with those without diabetes (2). Insulin resistance and hyperinsulinemia that occur in early diabetes have been proposed as one of the major mechanisms. Insulin stimulates growth of colonic epithelial and carcinoma cells in vitro and increases bioactive insulin-like growth factor 1 that can promote cell proliferation and inhibit apoptosis (3).

In contrast to the ample evidence linking diabetes to colorectal cancer development, only a few prospective cohort studies have been conducted to examine the association of diabetes with survival among patients with colorectal cancer (4, 5), leaving several questions unresolved. First, studies found shorter survival for patients with colorectal cancer and diabetes, but it is unclear whether these patients had increased mortality from the colorectal cancer itself or from other diseases associated with diabetes (e.g., other malignancies and cardiovascular disease). Second, with improving screening and treatment programs, a large fraction of patients with colorectal cancer is achieving long-term survival, with a 5-year survival rate of more than 60% (1). However, few studies have delineated the association of diabetes with survival by time period after colorectal cancer diagnosis. Third, the natural history of diabetes begins with insulin resistance and hyperinsulinemia, followed by hypoinsulinemia in later stages due to β cell dysfunction; the association of diabetes duration with colorectal cancer survival remains unclear.

To address these questions, we examined the association of diabetes with overall and cause-specific mortality among 4,038 participants diagnosed with colorectal cancer from two prospective U.S. cohorts, the Nurses' Health Study (NHS) and the Health Professionals Follow-Up Study (HPFS).

NHS was initiated in 1976 when 121,700 U.S. female nurses ages 30–55 years completed a mailed questionnaire describing demographics, lifestyle choices, and medical history (6). HPFS was initiated in 1986 when 51,529 U.S. men ages 40–75 years working in health professions completed a mailed questionnaire on health-related behaviors and medical history (7). Participants have provided updated information through biennial follow-up questionnaires. A high follow-up rate of more than 90% has been achieved in both cohorts.

This study included 4,038 participants with pathologically confirmed incident colorectal cancer diagnosed between 1976 and 2014. When a participant (or next of kin for decedents) reported a diagnosis of colorectal cancer on a follow-up questionnaire, we asked permission to obtain hospital records and pathology reports. For nonrespondents, the National Death Index was used to ascertain any diagnosis of colorectal cancer that contributed to death; we then asked permission from next of kin for decedents to obtain medical records. Study physicians reviewed these records to confirm the diagnosis and record information on important tumor characteristics. We estimate that 96%–97% of patients were identified through these methods (8, 9).

The study protocol was approved by the Institutional Review Boards of the Brigham and Women's Hospital (Boston, MA) and the Harvard T.H. Chan School of Public Health (Boston, MA), and those of participating registries as required. All participants provided written informed consent for the researchers to access their medical records. The study was conducted in concordance with the International Ethical Guidelines for Biomedical Research Involving Human Subjects (Council for International Organizations of Medical Sciences).

On baseline and biennial follow-up questionnaires, participants were asked whether they had ever been diagnosed with diabetes by a physician. To verify the diagnosis, a supplementary questionnaire was subsequently sent to obtain details on the date of diagnosis, symptoms, diagnostic tests, and treatment. In both cohorts, the validity of the supplementary questionnaire to confirm self-reported diabetes was established by review of medical records (10, 11).

In this study, diabetes status at colorectal cancer diagnosis was determined from participant report on biennial questionnaires together with the supplementary questionnaire. For those without a reported date of diabetes diagnosis, we used the return date of the biennial questionnaire on which they first reported being told by a physician that they had diabetes. Patients with diabetes were defined as those diagnosed with diabetes before or in the same month as colorectal cancer diagnosis; those with diabetes after colorectal cancer diagnosis were considered patients without diabetes. Categories of diabetes treatment included metformin, other oral medication, insulin injection, and neither oral medication nor insulin injection. Participants receiving metformin included those who used both metformin and other oral medication. Participants receiving insulin injection included those who initially used oral medication and switched to insulin injection.

Death ascertainment included reporting by family or postal authorities, and searching for names of persistent nonrespondents in the National Death Index, which has been shown to capture approximately 98% of deaths (12). Participants not found to be deceased were assumed to be alive. Cause of death was assigned by investigators blinded to other data according to the International Classification of Diseases, Revision 8. The primary outcome was overall mortality, and the secondary outcomes included mortality from colorectal cancer, other primary malignancies, and cardiovascular disease.

Cancer stage, grade of tumor differentiation, location of primary tumor, and year of diagnosis (as a surrogate for treatment) were extracted from medical records. Race and height were asked on the baseline questionnaire. Body weight, physical activity, cigarette smoking, and medication use (including dose and frequency of aspirin use) were surveyed every 2 years. Participants received additional questionnaires every 4 years to report their usual diet (including alcohol intake). In this study, data on body weight, physical activity, dietary intakes, and aspirin use were obtained from the questionnaire returned before colorectal cancer diagnosis. The Alternate Healthy Eating Index-2010 (AHEI-2010; ref. 13) was computed to measure overall diet quality. Regular aspirin users were defined as those who used aspirin at least two times per week, including standard and low-dose aspirin.

Follow-up time was calculated from colorectal cancer diagnosis to death or the end of follow-up (June 2014 for NHS and January 2014 for HPFS), whichever came first. Patients without diabetes were the main reference group for all analyses. The Kaplan–Meier method was used to calculate the 5-year survival rate and generate survival curves, with statistical significance evaluated by the log-rank test. Cox proportional hazards regression was used to calculate HRs and 95% CIs for overall and cause-specific mortality. In analyses with cause-specific mortality as the outcome, death from other causes was censored. To examine whether diabetes status can predict survival independent of tumor characteristics, the models were stratified by sex and cancer stage and adjusted for age at cancer diagnosis, grade of tumor differentiation, location of primary tumor, and year of diagnosis. To control for the lifestyle factors that may affect survival, the models were additionally adjusted for body mass index, physical activity, smoking status, alcohol intake, AHEI-2010, and regular aspirin use.

We tested whether the association of diabetes with overall and cause-specific mortality changed over time, by evaluating the product of diabetes status and follow-up time as a continuous variable. The assumption was violated for overall mortality (P = 0.009 in women and P = 0.02 in men), which was addressed by including in the model interaction terms between diabetes status and follow-up periods, including ≤5, >5–10, and >10 years (the assumption was satisfied within each follow-up period with P ≥ 0.46). In contrast, we did not find evidence that the association of diabetes with any cause-specific mortality changed over time (P ≥ 0.47).

Heterogeneity by sex was tested by Cochran Q statistic (14). We examined whether the association of diabetes with 5-year overall mortality and colorectal cancer–specific mortality differed by age at cancer diagnosis, body mass index, cancer stage, grade of tumor differentiation, and location of primary tumor. These analyses were conducted by entering the cross-product of diabetes status and the potential effect modifier in the model, evaluated by the likelihood ratio test. All analyses were performed with SAS Software, version 9.4 (SAS Institute). All P values are two-sided.

Among 4,038 patients with colorectal cancer (65.2% female), 429 (10.6%) had diabetes at cancer diagnosis. Baseline characteristics according to diabetes status are shown in Table 1. Compared with those without diabetes, patients with colorectal cancer and diabetes were older, had a higher body mass index, were less physically active, consumed less alcohol, and were less likely to be current smokers (particularly in women), and more likely to use aspirin.

The median follow-up among patients who were alive at the end of follow-up was 12.6 years (interquartile range, 6.6–18.6 years). During the follow-up period, we documented 2,440 deaths, 1,450 (59.4%) of which were due to colorectal cancer. Non-colorectal cancer causes of death included other primary malignancies (n = 244), cardiovascular disease (n = 219), neurologic disorders (n = 113), respiratory disease (n = 90), cerebrovascular disease (n = 78), and other or unknown reasons (n = 246). The 244 deaths from other malignancies included 59 deaths from lung cancer, 37 from hematologic cancers, 36 from other gastrointestinal cancers, 36 from genitourinary cancers, 28 from breast cancer, eight from skin cancer, five from brain cancer, and 35 from cancers of other or unspecified sites. A summary of causes of death according to follow-up period is presented in Supplementary Table S1. During ≤5, >5–10, and >10 years after diagnosis, colorectal cancer accounted for 83.8%, 33.6%, and 11.1% of deaths, respectively.

The overall survival curves by diabetes status are shown in Fig. 1 (log-rank P < 0.0001 in women and P = 0.09 in men). The median overall survival for patients with and without diabetes was 6.3 and 11.1 years in women, and 7.8 and 8.8 years in men, respectively (Table 2). In the first 5 years after colorectal cancer diagnosis, diabetes was associated with increased overall mortality in women (HR, 1.22; 95% CI, 1.00–1.49), but not in men (HR, 0.83; 95% CI, 0.62–1.12; P heterogeneity by sex = 0.04; Table 2). The association between diabetes and 5-year overall mortality did not differ significantly by age at cancer diagnosis, cancer stage, grade of tumor differentiation, location of primary tumor, or body mass index (Supplementary Table S2). Beyond 5 years, diabetes was associated with increased overall mortality with no evidence of sex heterogeneity (P_{heterogeneity} ≥ 0.50); in women and men combined, the HRs were 1.45 (95% CI, 1.09–1.93) during >5–10 years and 2.58 (95% CI, 1.91–3.50) during >10 years (Table 2).

The association between diabetes and cause-specific mortality did not differ significantly by sex (P_{heterogeneity} = 0.10, 0.28, and 0.88 for mortality from colorectal cancer, other malignancies, and cardiovascular disease, respectively; Table 3). In women and men combined, diabetes was not associated with colorectal cancer–specific mortality (HR, 1.09; 95% CI, 0.91–1.30), but associated with increased mortality from other malignancies (HR, 1.78; 95% CI, 1.18–2.67) and cardiovascular disease (HR, 1.93; 95% CI, 1.29–2.91). Nonetheless, a nonsignificant increase in colorectal cancer–specific mortality was noted for women with diabetes (HR, 1.21; 95% CI, 0.98–1.49). The association between diabetes and colorectal cancer–specific mortality did not differ significantly by age at cancer diagnosis, cancer stage, grade of tumor differentiation, location of primary tumor, or body mass index (Supplementary Table S3).

We analyzed overall and cause-specific mortality by diabetes duration at colorectal cancer diagnosis (≤10 or >10 years) in women and men separately, and found increased mortality with longer diabetes duration in women. Compared with those without diabetes, women with diabetes for more than 10 years had increased overall mortality during all follow-up periods, with HRs of 1.39 (95% CI, 1.08–1.78) during ≤5 years, 2.30 (95% CI, 1.39–3.78) during >5–10 years, and 4.21 (95% CI, 2.27–7.79) during >10 years (Table 4). These women also had increased mortality from colorectal cancer (HR, 1.33; 95% CI, 1.01–1.76), other malignancies (HR, 2.66; 95% CI, 1.39–5.09), and cardiovascular disease (HR, 2.54; 95% CI, 1.09–5.89; Supplementary Table S4).

In men with colorectal cancer, a clear association with diabetes duration was not evident. Men with diabetes for more than 10 years did not have a significant increase in overall or any cause-specific mortality, compared with those without diabetes (Table 4; Supplementary Table S4).

Although somewhat limited by sample size, we analyzed 5-year overall and colorectal cancer–specific mortality by diabetes treatment in women and men combined (Supplementary Table S5). Compared with those without diabetes, those with diabetes receiving metformin had decreased 5-year overall mortality (HR, 0.52; 95% CI, 0.30–0.91), and those receiving neither oral medication nor insulin injection had increased 5-year overall mortality (HR, 1.68; 95% CI, 1.17–2.41). Similar associations with colorectal cancer–specific mortality were noted.

Among the 4,038 patients with colorectal cancer from two prospective U.S. cohorts, preexisting diabetes was associated with a modest increase in overall mortality in women, but not in men, during the first 5 years after cancer diagnosis. Beyond 5 years, diabetes was associated with substantially increased overall mortality with no evidence of sex heterogeneity. The excess mortality was largely caused by cardiovascular disease and malignancies other than colorectal cancer. Only women with diabetes for more than 10 years had increased mortality from colorectal cancer.

To date, many studies have evaluated the association between preexisting diabetes and survival among patients with colorectal cancer. The vast majority of these studies were hospital based and enrolled a small number of patients. Only a few population-based studies have been conducted (15), most of which identified patients with colorectal cancer from cancer registries or primary care data. These studies generally lacked information on lifestyle factors that have been associated with colorectal cancer survival, such as excess body weight, cigarette smoking, alcohol intake, and aspirin use (16). Given the link between unhealthy lifestyle behaviors and diabetes, multivariate modeling is necessary to evaluate the independent association between diabetes and patient survival. Another issue in these studies relates to the completeness and accuracy of the linked databases to ascertain diabetes status, such as the Medicare claims data that were shown to capture only 69% of people with diabetes (17).

To our knowledge, only two prospective cohort studies have been conducted to examine the association between diabetes and survival among patients with all stages of colorectal cancer with adjustment for lifestyle factors (4, 5). Among 2,278 patients with colorectal cancer from the Cancer Prevention Study-II Nutrition Cohort, those with diabetes had increased risk of overall mortality (HR, 1.53; 95% CI, 1.23–1.83), colorectal cancer–specific mortality (HR, 1.29; 95% CI, 0.98–1.70), and cardiovascular disease–specific mortality (HR, 2.16; 95% CI, 1.44–3.24; ref. 4). Among 3,913 patients within the Multiethnic Cohort Study in California and Hawaii, only diabetes that existed for at least 10 years was associated with overall (HR, 1.49; 95% CI, 1.22–1.82) and colorectal cancer–specific mortality (HR, 1.48; 95% CI, 1.06–2.07; ref. 5).

This study found a time-varying association between diabetes and overall mortality after colorectal cancer diagnosis. In the first 5 years, women with colorectal cancer and diabetes had an estimated 22% higher risk of overall mortality, but the mortality was not elevated in men. Beyond 5 years, diabetes was associated with a much greater increase in overall mortality with no evidence of sex heterogeneity. One explanation is that diabetes was more strongly associated with mortality from non-colorectal cancer diseases, such as other malignancies and cardiovascular disease, which accounted for the majority of deaths beyond 5 years. Another factor was diabetes duration that increased with follow-up and was positively associated with mortality in this study.

We found increased colorectal cancer–specific mortality for women with diabetes for more than 10 years. If hyperinsulinemia that occurs in the early stage of diabetes accounts for this association, we would expect a similar or greater increase in the mortality among those with diabetes for a shorter period, which we did not find. Moreover, a previous study conducted in the same cohorts found no association between diabetes duration and risk for colorectal cancer development (18). Together, these data do not support a strong role of hyperinsulinemia in the relationship between diabetes and colorectal carcinogenesis.

The somewhat stronger association between diabetes and mortality in female patients with colorectal cancer is a novel finding and requires confirmation in future studies. One possible explanation is that glycemic control for diabetics is generally worse in women than in men (19). Studies also found that women were more likely to receive less aggressive treatment for diabetes (20) and had a lower adherence to glucose-lowering medication (21) than men. Hyperglycemia can contribute to cancer progression by affecting cancer cell behaviors, including proliferation, apoptosis, migration, and invasion, and by enhancing chemotherapy resistance and intolerance (22, 23). The negative influence of hyperglycemia on colorectal cancer survival was supported by our observation that patients with colorectal cancer and diabetes who used neither insulin injection nor oral medication for glycemic control had the worst survival.

Our observation that colorectal cancer survivors with diabetes had increased mortality from other malignancies has not been reported previously, but is biologically plausible. Diabetes has been associated with increased risk (24, 25) and decreased survival (26–28) of several cancers (e.g., liver, pancreas, endometrium, breast, and prostate). Notably, the nearly 1.8 times higher risk of mortality from other malignancies is more pronounced than the association between diabetes and cancer mortality in the general population with no colorectal cancer (typically <1.3 times; refs. 29–31). Colorectal cancer development is known to be associated with multiple diet and lifestyle factors, which also contribute to morbidity and mortality of other cancers (32) and may account for the greater association between diabetes and cancer mortality among patients with colorectal cancer. Regardless of the underlying mechanisms, our findings suggest that prevention of other malignancies should be an important concern for colorectal cancer survivors with diabetes.

Consistent with previous reports (4, 33), we found that patients with colorectal cancer and diabetes had substantially increased mortality from cardiovascular disease. Diabetes and cardiovascular disease are closely linked, as several of the risk factors for cardiovascular disease (e.g., obesity, hypertension, hyperglycemia, and dyslipidemia) are common in people with diabetes (34). In the general population, cardiovascular disease is the major cause of morbidity and mortality among people with diabetes, so reducing risk factors for cardiovascular disease is a critical part of diabetes management. While this study included patients diagnosed with colorectal cancer as early as the 1970s, recent reports demonstrate that owing to effective treatment for diabetes, there has been a substantial decline in cardiovascular disease mortality among people with diabetes over the past decades (35, 36).

This study has several strengths, including the prospective design, long follow-up time, high follow-up rates, biennial self-report on diabetes status and subsequent confirmation, as well as detailed information on tumor characteristics and lifestyle factors. Limitations of this study also require consideration. We did not control for differences in colorectal cancer treatment, because this information was not systematically collected in our cohorts. We cannot exclude that patients with colorectal cancer and diabetes may have been treated less aggressively for colorectal cancer (37) and, therefore, had worse survival. Our study participants were predominantly White, and studies in other racial populations are warranted.

In conclusion, our study suggests that among patients with colorectal cancer, preexisting diabetes is associated with increased risk of long-term mortality, particularly from other malignancies and cardiovascular disease. These findings highlight the importance of cardioprotection and prevention of other malignancies to colorectal cancer survivors with diabetes.

B.M. Wolpin reports grants and personal fees from Celgene, grants from Eli Lilly, and personal fees from GRAIL and BiolineRx outside the submitted work. J.A. Meyerhardt reports personal fees from COTA Healthcare, Taiho Pharmaceutical, and Ignyta outside the submitted work. A.T. Chan reports grants and personal fees from Bayer Pharma AG and personal fees from Pfizer Inc. and Boehringer Ingelheim outside the submitted work. C.S. Fuchs reports personal fees from Agios, Amylin Pharma, Bain Capital, CytomX Therapeutics, Daiichi-Sankyo, Eli Lilly, Entrinsic Health, EvolveImmune Therapeutics, Genentech, Merck, Taiho, Unum Therapeutics, and AstraZeneca outside the submitted work, and serves as a director for CytomX Therapeutics, owns unexercised stock options for CytomX and Entrinsic Health, is a cofounder of EvolveImmune Therapeutics and has equity in this private company, and has provided expert testimony for Amylin Pharmaceuticals and Eli Lilly. S. Ogino reports grants from NIH during the conduct of the study. K. Ng reports grants from NCI, Pussycat Foundation, Project P Fund, and Dana-Farber Cancer Institute during the conduct of the study, as well as grants from Department of Defense, Cancer Research UK, Janssen, Revolution Medicines, Genentech, and Gilead Sciences; nonfinancial support from Pharmavite and Evergrande Group; and personal fees from Bayer, Seattle Genetics, Array Biopharma, BiomX, and X-Biotix Therapeutics outside the submitted work. No disclosures were reported by the other authors.

C. Yuan: Conceptualization, data curation, formal analysis, investigation, methodology, writing–original draft, writing–review and editing. X. Zhang: Conceptualization, data curation, formal analysis, investigation, methodology, writing–review and editing. A. Babic: Writing–review and editing. V. Morales-Oyarvide: Writing–review and editing. Y. Zhang: Writing–review and editing. S.A. Smith-Warner: Methodology, writing–review and editing. K. Wu: Methodology, writing–review and editing. M. Wang: Methodology, writing–review and editing. B.M. Wolpin: Writing–review and editing. J.A. Meyerhardt: Data curation, writing–review and editing. A.T. Chan: Writing–review and editing. F.B. Hu: Writing–review and editing. C.S. Fuchs: Writing–review and editing. S. Ogino: Supervision, funding acquisition, investigation, writing–review and editing. E.L. Giovannucci: Conceptualization, data curation, formal analysis, supervision, investigation, methodology, writing–original draft, writing–review and editing. K. Ng: Conceptualization, data curation, formal analysis, supervision, funding acquisition, investigation, methodology, writing–original draft, writing–review and editing.

We would like to thank the participants and staff of the Nurses' Health Study and the Health Professionals Follow-Up Study for their valuable contributions, as well as the following state cancer registries for their help: AL, AZ, AR, CA, CO, CT, DE, FL, GA, ID, IL, IN, IA, KY, LA, ME, MD, MA, MI, NE, NH, NJ, NY, NC, ND, OH, OK, OR, PA, RI, SC, TN, TX, VA, WA, WY. The authors assume full responsibility for analyses and interpretation of these data. The Nurses' Health Study is supported by the NIH grants UM1 CA186107 and P01 CA87969. The Health Professionals Follow-Up Study is supported by the NIH grant U01 CA167552. This work was additionally supported by the Pussycat Foundation Helen Gurley Brown Presidential Initiative (to C. Yuan and K. Ng), by the NIH grant K07 CA188126 and American Cancer Society Research Scholar grant RSG NEC-130476 (to X. Zhang), by the NIH grants R01 CA137178 and K24 DK098311 and the Damon Runyon Cancer Research Foundation (to A.T. Chan), by the NIH grant P50 CA127003 (to C.S. Fuchs), by the NIH grant R35 CA197735 (to S. Ogino), by the NIH grants K07 CA148894 and R01 CA205406 and the Project P Fund (to K. Ng), and by the Entertainment Industry Foundation's National Colorectal Cancer Research Alliance (NCCRA).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.