https://orcid.org/0000-0003-2553-6170
Abstract
Although diabetes mellitus (DM) was reported to be associated with incident colorectal cancer (CRC), the detailed association between fasting plasma glucose (FPG) and incident CRC has not been fully understood.
We assessed whether hyperglycemia is associated with a higher risk for CRC.
Analyses were conducted using the JMDC Claims Database [n = 1 441 311; median age (interquartile range), 46 (40-54) years; 56.6% men). None of the participants were taking antidiabetic medication or had a history of CRC, colorectal polyps, or inflammatory bowel disease. Participants were categorized as normal FPG (FPG level < 100 mg/dL; 1 125 647 individuals), normal-high FPG (FPG level = 100-109 mg/dL; 210 365 individuals), impaired fasting glucose (IFG; FPG level = 110-125 mg/dL; 74 836 individuals), and DM (FPG level ≥ 126 mg/dL; 30 463 individuals).
Over a mean follow-up of 1137 ± 824 days, 5566 CRC events occurred. After multivariable adjustment, the hazard ratios for CRC events were 1.10 (95% CI 1.03-1.18) for normal-high FPG, 1.24 (95% CI 1.13-1.37) for IFG, and 1.36 (95% CI 1.19-1.55) for DM vs normal FPG. We confirmed this association in sensitivity analyses excluding those with a follow-up of< 365 days and obese participants.
The risk of CRC increased with elevated FPG category. FPG measurements would help to identify people at high-risk for future CRC.
Colorectal cancer (CRC) is the third most commonly diagnosed cancer and the second leading cause of cancer-related deaths worldwide, with an estimated 1.8 million new cases and 861 000 deaths annually according to the GLOBOCAN data (1). In the United States, approximately 150 000 new CRC cases are diagnosed, and approximately 54 000 Americans die annually because of CRC (2). Epidemiological data suggest that hyperglycemia or diabetes mellitus (DM) is associated with a higher incidence and mortality of CRC (3-6). A recent meta-analysis of six studies with 2 969 306 participants and 62 814 CRC cases showed that the relative risk of CRC per 20 mg/dL increase in fasting plasma glucose (FPG) was 1.015 (95% CI 1.012-1.019) (7). However, there are several limitations in the preceding studies that focus on the association between FPG and incident CRC. First, shared risk factors (including obesity, high waist circumference, and hypertension) that could also affect the incidence of CRC may confound the relationship between FPG and cancer (8-12). Second, prior studies included individuals taking antidiabetic medications. Third, preceding studies mainly compared the risk of CRC between DM and non-DM or between hyperglycemia and non-hyperglycemia, and therefore, it has not been fully understood whether the risk of CRC would increase stepwise with increasing FPG. Fourth, most of the aforementioned studies included individuals at a high risk of CRC, including colorectal polyps, Crohn’s disease, and ulcerative colitis. Furthermore, the estimated proportion of CRC that would be preventable if FPG could be normalized has not yet been evaluated. Using data from the health claims database of the JMDC that excluded individuals with prior CRC, colorectal polyps, and inflammatory bowel disease (13-16), we assessed whether subjects with higher FPG are at a higher risk of incident CRC events than those with normal FPG. We also assessed the proportion of incident CRCs that would be potentially preventable when FPG could be lowered to the normal range (<100 mg/dL) using the relative risk reduction (RRR).
Materials and Methods
Study Design and Data Source
We conducted this retrospective observational study using the JMDC Claims Database (JMDC Inc., Tokyo, Japan) (17) between January 2005 and August 2018. The Japanese government provides a universal health insurance program for all registered inhabitants, and each employer is obliged by law to provide to its employees an opportunity of health check-up. Medical and pharmacy claims data combined with health check-up data from employees’ health insurance programs were obtained in an anonymous format from the JMDC Claims Database. The JMDC Claims Database includes individual health insurance claims records from more than 60 insurers. The JMDC Claims Database includes information on employees, including demographics, prior medical history, medication status, and hospital claims records with International Classification of Diseases, 10th revision (ICD-10) coding. For the current analyses, we selected the records of individuals (n = 1 987 819) who underwent assessments for FPG. We excluded individuals taking antidiabetic medications (n = 59 189); those aged <20 years (n = 10 032); those with a history of colorectal disease, including CRC (ICD-10: C18, C19, and C20), colorectal polyps (ICD-10: K635, K621), ulcerative colitis (ICD-10: K51), and Crohn’s disease (ICD-10: K50) (n = 40 885); and any missing values on body mass index (BMI) (n = 768), waist circumference (n = 80 931), medications for hypertension or dyslipidemia (n = 134), blood pressure (n = 1 457), low-density lipoprotein cholesterol (n = 726), high-density lipoprotein cholesterol (n = 60), triglycerides (n = 397), cigarette smoking (n = 6300), alcohol consumption (n = 236 709), and physical inactivity (n = 108 920). A flowchart defining the sample used in the analyses is shown in Figure 1. After all exclusion criteria were applied, data from 1 441 311 individuals were analyzed in this study.
Flowchart. We selected records of individuals (n = 1 987 819) who underwent assessments of FPG. We excluded individuals taking antidiabetic medications (n = 59 189); those aged <20 years (n = 10 032); and those with history of colorectal disease including CRC (C18, C19, C20), colorectal polyp (K635, K621), ulcerative colitis (K51), and Crohn’s disease (K50) (n = 40 885); and those any missing data on body mass index (n = 768), waist circumference (n = 80 931), medications for hypertension or dyslipidemia (n = 134), blood pressure (n = 1 457), low-density lipoprotein cholesterol (n = 726), high-density lipoprotein cholesterol (n = 60), triglycerides (n = 397), cigarette smoking (n = 6300), alcohol consumption (n = 236 709), and physical inactivity (n = 108 920). Finally, we analyzed 1 441 311 participants in this study.
Ethics
This study was conducted according to the ethical guidelines of our institution (approval by the Ethical Committee of The University of Tokyo: 2018-10862) and in accordance with the principles of the Declaration of Helsinki. The requirement for informed consent was waived because all data in the JMDC Claims Database were anonymized and de-identified. All data were compliant with the International Conference on Harmonization guidelines (18).
Fasting Plasma Glucose Category and Other Measurements
Health check-up data, including BMI, waist circumference, history of hypertension, DM, dyslipidemia, blood pressure, and fasting laboratory values, including FPG, were collected using standardized protocols across study centers. We categorized the study population into 4 groups based on FPG at the initial health check-up; normal FPG, defined as an FPG level <100 mg/dL; normal-high FPG and IFG, defined as 109 mg/dL ≥ FPG level ≥ 100 mg/dL and 125 mg/dL ≥ FPG level ≥ 110 mg/dL, respectively; and DM defined as an FPG level ≥ 126 mg/dL (19,20). Obesity was defined as BMI ≥ 25 kg/m2 (15,20). High waist circumference was defined as waist circumference ≥ 85 cm for men and ≥90 cm for women (21,22). Hypertension was defined as systolic blood pressure ≥ 140 mmHg, diastolic blood pressure ≥ 90 mmHg or use of antihypertensive medications (20,21). Dyslipidemia was defined as low-density lipoprotein cholesterol level ≥ 140 mg/dL, high-density lipoprotein cholesterol level <40 mg/dL, triglycerides ≥ 150 mg/dL, or use of lipid-lowering medications (20,21). Information on cigarette smoking (current or noncurrent) and alcohol consumption (daily or occasional/none) was self-reported. As previously described (14), we defined physical inactivity as applying to individuals who did not report engaging in 30 min of exercise more than 2 times a week or walking ≥ 1 h per day.
Outcomes
Outcomes data were collected between January 2005 and August 2018. The primary outcomes included CRC (ICD-10: C18, C19, and C20).
Statistical Analysis
Summary statistics for the characteristics of the participants in the FPG groups were calculated. The statistical significance of differences among groups was determined using analysis of variance for continuous variables and chi-squared tests for categorical variables. The cumulative incidence of CRC events in the FPG groups was calculated using the Kaplan-Meier method. We conducted Cox regression analyses to identify the association between the FPG category and the subsequent risk of CRC. Hazard ratios (HRs) were calculated in an unadjusted model and after adjustment for potential confounders, including age, sex, obesity, high waist circumference, hypertension, dyslipidemia, current cigarette smoking, alcohol consumption (daily), and physical inactivity. We conducted 7 sensitivity analyses. First, we analyzed the relationship between FPG as a continuous variable and incident CRC. To detect any possible linear or non-linear dependency in regression models and to allow for flexible interpretation of the relationship between continuous covariates and study outcomes, continuous changes in FPG were assessed through shape-restricted cubic spline regression models with knots at 100 mg/dL, 110 mg/dL, and 126 mg/dL, with 100 mg/dL as the reference (23). Second, we performed multiple imputation for dealing with missing data, as previously described (24,25). We assumed that missing data for covariates occurred independently of FPG and imputed missing data using the chained equation method with 20 iterations as described by Aloisio et al (26). We obtained HRs and standard errors using Rubin’s rules (27). Third, to minimize the possible influence of latent CRC or reverse causation, we excluded the participants with follow-up period for CRC shorter than 1 year. Fourth, we excluded the participants who had obesity because obesity is known to be associated with the incidence of CRC (8,28). Fifth, because the association of hyperglycemia with CRC may differ by sex (29,30), we conducted stratified analyses by sex. The P-values for interactions between groups were also calculated. Sixth, we excluded from the study population participants diagnosed with CRC but no treatment history confirmed. We defined colon resection (K719), colorectal mucosal resection (K721), rectal resection (K740), or others (K726, K728, K732, and K736) as surgery for CRC and use of fluorouracil, irinotecan, oxaliplatin, or capecitabine as chemotherapy for CRC. Seventh, we created a directed acyclic graph to present the potential causal relationship (31). We estimated RRR as the proportion of CRC events that would be preventable if the FPG of those with normal-high FPG, IFG, and DM were lowered to normal levels based on HRs of normal-high, IFG, and DM for CRC event (32,33). Statistical significance was set at P < 0.05. All statistical analyses were performed using the SPSS software version 25 and STATA version 16.
Results
Characteristics of Study Population
The characteristics of the study participants (n = 1 441 311) are presented in Table 1. The median (interquartile range) age was 46 (40-54) years, and 815 212 participants (56.6%) were men. Using FPG measurements at baseline, participants were categorized as having normal FPG (n = 1 125 647), normal-high FPG (n = 210 365), IFG (n = 74 836), or DM (n = 30 463). Participants with normal-high FPG, IFG, and DM were older and more likely to be men, current smokers, and habitual drinkers than their counterparts with normal FPG. Participants in the normal-high FPG, IFG, and DM groups had higher BMI, waist circumference, blood pressure, serum low-density lipoprotein-cholesterol, and triglyceride levels and lower serum high-density lipoprotein-cholesterol levels than their counterparts in the normal FPG group.
Clinical characteristics of study population
| Variable | Normal FPG (n = 1 125 647) | Normal-high FPG (n = 210 365) | Impaired fasting glucose (n = 74 836) | Diabetes mellitus (n = 30 463) | P-value |
|---|---|---|---|---|---|
| Glucose, mg/dL | 89 (84-94) | 103 (101-106) | 114 (112-118) | 140 (131-161) | — |
| Age, years | 45 (40-52) | 51 (44-58) | 54 (47-60) | 54 (47-60) | <0.001 |
| Male sex, n (%) | 583 344 (51.8) | 149 754 (71.2) | 57 640 (77.0) | 24 474 (80.3) | <0.001 |
| Body mass index, kg/m2 | 21.9 (19.9-24.2) | 23.6 (21.5-25.9) | 24.4 (22.2-27.0) | 25.3 (22.9-28.3) | <0.001 |
| Obesity, n (%) | 214 440 (19.1) | 71 534 (34.0) | 32 950 (44.0) | 16 351 (53.7) | <0.001 |
| Waist circumference, cm | 79 (73-85) | 84 (78-90) | 86 (81-93) | 89 (83-96) | <0.001 |
| High waist circumference, n (%) | 249 314 (22.1) | 86 466 (41.1) | 39 127 (52.3) | 19 020 (62.4) | <0.001 |
| Hypertension, n (%) | 162 729 (14.5) | 65 921 (31.3) | 33 936 (45.3) | 15 648 (51.4) | <0.001 |
| Systolic blood pressure, mmHg | 116 (106-126) | 124 (114-134) | 128 (118-139) | 131 (120-143) | <0.001 |
| Diastolic blood pressure, mmHg | 71 (64-80) | 78 (70-85) | 80 (73-88) | 82 (74-90) | <0.001 |
| Dyslipidemia, n (%) | 401 053 (35.6) | 113 046 (53.7) | 47 129 (63.0) | 21 559 (70.8) | <0.001 |
| Low-density lipoprotein cholesterol, mg/dL | 117 (97-138) | 126 (106-148) | 128 (108-150) | 131 (108-154) | <0.001 |
| High-density lipoprotein cholesterol, mg/dL | 63 (53-75) | 59 (49-70) | 56 (48-68) | 53 (45-63) | <0.001 |
| Triglyceride, mg/dL | 77 (55-114) | 101 (71-148) | 115 (81-169) | 137 (93-206) | <0.001 |
| Cigarette smoking | 267 538 (23.8) | 56 675 (26.9) | 21 420 (28.6) | 10 749 (35.3) | <0.001 |
| Alcohol consumption | 241 643 (21.5) | 70 246 (33.4) | 28 038 (37.5) | 10 293 (33.8) | <0.001 |
| Physical inactivity | 622 355 (55.3) | 115 607 (55.0) | 41 290 (55.2) | 17 340 (56.9) | <0.001 |
| Variable | Normal FPG (n = 1 125 647) | Normal-high FPG (n = 210 365) | Impaired fasting glucose (n = 74 836) | Diabetes mellitus (n = 30 463) | P-value |
|---|---|---|---|---|---|
| Glucose, mg/dL | 89 (84-94) | 103 (101-106) | 114 (112-118) | 140 (131-161) | — |
| Age, years | 45 (40-52) | 51 (44-58) | 54 (47-60) | 54 (47-60) | <0.001 |
| Male sex, n (%) | 583 344 (51.8) | 149 754 (71.2) | 57 640 (77.0) | 24 474 (80.3) | <0.001 |
| Body mass index, kg/m2 | 21.9 (19.9-24.2) | 23.6 (21.5-25.9) | 24.4 (22.2-27.0) | 25.3 (22.9-28.3) | <0.001 |
| Obesity, n (%) | 214 440 (19.1) | 71 534 (34.0) | 32 950 (44.0) | 16 351 (53.7) | <0.001 |
| Waist circumference, cm | 79 (73-85) | 84 (78-90) | 86 (81-93) | 89 (83-96) | <0.001 |
| High waist circumference, n (%) | 249 314 (22.1) | 86 466 (41.1) | 39 127 (52.3) | 19 020 (62.4) | <0.001 |
| Hypertension, n (%) | 162 729 (14.5) | 65 921 (31.3) | 33 936 (45.3) | 15 648 (51.4) | <0.001 |
| Systolic blood pressure, mmHg | 116 (106-126) | 124 (114-134) | 128 (118-139) | 131 (120-143) | <0.001 |
| Diastolic blood pressure, mmHg | 71 (64-80) | 78 (70-85) | 80 (73-88) | 82 (74-90) | <0.001 |
| Dyslipidemia, n (%) | 401 053 (35.6) | 113 046 (53.7) | 47 129 (63.0) | 21 559 (70.8) | <0.001 |
| Low-density lipoprotein cholesterol, mg/dL | 117 (97-138) | 126 (106-148) | 128 (108-150) | 131 (108-154) | <0.001 |
| High-density lipoprotein cholesterol, mg/dL | 63 (53-75) | 59 (49-70) | 56 (48-68) | 53 (45-63) | <0.001 |
| Triglyceride, mg/dL | 77 (55-114) | 101 (71-148) | 115 (81-169) | 137 (93-206) | <0.001 |
| Cigarette smoking | 267 538 (23.8) | 56 675 (26.9) | 21 420 (28.6) | 10 749 (35.3) | <0.001 |
| Alcohol consumption | 241 643 (21.5) | 70 246 (33.4) | 28 038 (37.5) | 10 293 (33.8) | <0.001 |
| Physical inactivity | 622 355 (55.3) | 115 607 (55.0) | 41 290 (55.2) | 17 340 (56.9) | <0.001 |
Data are expressed as median (interquartile range) or number (percentage). P-values were calculated using the analysis of variance for continuous variables and chi-square tests for categorical variables. Participants were categorized based on fasting plasma glucose (FPG) category as normal FPG (FPG level < 100 mg/dL), normal-high FPG (FPG level of 100-109 mg/dL), impaired fasting glucose (FPG level of 110-125 mg/dL), and diabetes mellitus (FPG level ≥ 126 mg/dL).
Clinical characteristics of study population
| Variable | Normal FPG (n = 1 125 647) | Normal-high FPG (n = 210 365) | Impaired fasting glucose (n = 74 836) | Diabetes mellitus (n = 30 463) | P-value |
|---|---|---|---|---|---|
| Glucose, mg/dL | 89 (84-94) | 103 (101-106) | 114 (112-118) | 140 (131-161) | — |
| Age, years | 45 (40-52) | 51 (44-58) | 54 (47-60) | 54 (47-60) | <0.001 |
| Male sex, n (%) | 583 344 (51.8) | 149 754 (71.2) | 57 640 (77.0) | 24 474 (80.3) | <0.001 |
| Body mass index, kg/m2 | 21.9 (19.9-24.2) | 23.6 (21.5-25.9) | 24.4 (22.2-27.0) | 25.3 (22.9-28.3) | <0.001 |
| Obesity, n (%) | 214 440 (19.1) | 71 534 (34.0) | 32 950 (44.0) | 16 351 (53.7) | <0.001 |
| Waist circumference, cm | 79 (73-85) | 84 (78-90) | 86 (81-93) | 89 (83-96) | <0.001 |
| High waist circumference, n (%) | 249 314 (22.1) | 86 466 (41.1) | 39 127 (52.3) | 19 020 (62.4) | <0.001 |
| Hypertension, n (%) | 162 729 (14.5) | 65 921 (31.3) | 33 936 (45.3) | 15 648 (51.4) | <0.001 |
| Systolic blood pressure, mmHg | 116 (106-126) | 124 (114-134) | 128 (118-139) | 131 (120-143) | <0.001 |
| Diastolic blood pressure, mmHg | 71 (64-80) | 78 (70-85) | 80 (73-88) | 82 (74-90) | <0.001 |
| Dyslipidemia, n (%) | 401 053 (35.6) | 113 046 (53.7) | 47 129 (63.0) | 21 559 (70.8) | <0.001 |
| Low-density lipoprotein cholesterol, mg/dL | 117 (97-138) | 126 (106-148) | 128 (108-150) | 131 (108-154) | <0.001 |
| High-density lipoprotein cholesterol, mg/dL | 63 (53-75) | 59 (49-70) | 56 (48-68) | 53 (45-63) | <0.001 |
| Triglyceride, mg/dL | 77 (55-114) | 101 (71-148) | 115 (81-169) | 137 (93-206) | <0.001 |
| Cigarette smoking | 267 538 (23.8) | 56 675 (26.9) | 21 420 (28.6) | 10 749 (35.3) | <0.001 |
| Alcohol consumption | 241 643 (21.5) | 70 246 (33.4) | 28 038 (37.5) | 10 293 (33.8) | <0.001 |
| Physical inactivity | 622 355 (55.3) | 115 607 (55.0) | 41 290 (55.2) | 17 340 (56.9) | <0.001 |
| Variable | Normal FPG (n = 1 125 647) | Normal-high FPG (n = 210 365) | Impaired fasting glucose (n = 74 836) | Diabetes mellitus (n = 30 463) | P-value |
|---|---|---|---|---|---|
| Glucose, mg/dL | 89 (84-94) | 103 (101-106) | 114 (112-118) | 140 (131-161) | — |
| Age, years | 45 (40-52) | 51 (44-58) | 54 (47-60) | 54 (47-60) | <0.001 |
| Male sex, n (%) | 583 344 (51.8) | 149 754 (71.2) | 57 640 (77.0) | 24 474 (80.3) | <0.001 |
| Body mass index, kg/m2 | 21.9 (19.9-24.2) | 23.6 (21.5-25.9) | 24.4 (22.2-27.0) | 25.3 (22.9-28.3) | <0.001 |
| Obesity, n (%) | 214 440 (19.1) | 71 534 (34.0) | 32 950 (44.0) | 16 351 (53.7) | <0.001 |
| Waist circumference, cm | 79 (73-85) | 84 (78-90) | 86 (81-93) | 89 (83-96) | <0.001 |
| High waist circumference, n (%) | 249 314 (22.1) | 86 466 (41.1) | 39 127 (52.3) | 19 020 (62.4) | <0.001 |
| Hypertension, n (%) | 162 729 (14.5) | 65 921 (31.3) | 33 936 (45.3) | 15 648 (51.4) | <0.001 |
| Systolic blood pressure, mmHg | 116 (106-126) | 124 (114-134) | 128 (118-139) | 131 (120-143) | <0.001 |
| Diastolic blood pressure, mmHg | 71 (64-80) | 78 (70-85) | 80 (73-88) | 82 (74-90) | <0.001 |
| Dyslipidemia, n (%) | 401 053 (35.6) | 113 046 (53.7) | 47 129 (63.0) | 21 559 (70.8) | <0.001 |
| Low-density lipoprotein cholesterol, mg/dL | 117 (97-138) | 126 (106-148) | 128 (108-150) | 131 (108-154) | <0.001 |
| High-density lipoprotein cholesterol, mg/dL | 63 (53-75) | 59 (49-70) | 56 (48-68) | 53 (45-63) | <0.001 |
| Triglyceride, mg/dL | 77 (55-114) | 101 (71-148) | 115 (81-169) | 137 (93-206) | <0.001 |
| Cigarette smoking | 267 538 (23.8) | 56 675 (26.9) | 21 420 (28.6) | 10 749 (35.3) | <0.001 |
| Alcohol consumption | 241 643 (21.5) | 70 246 (33.4) | 28 038 (37.5) | 10 293 (33.8) | <0.001 |
| Physical inactivity | 622 355 (55.3) | 115 607 (55.0) | 41 290 (55.2) | 17 340 (56.9) | <0.001 |
Data are expressed as median (interquartile range) or number (percentage). P-values were calculated using the analysis of variance for continuous variables and chi-square tests for categorical variables. Participants were categorized based on fasting plasma glucose (FPG) category as normal FPG (FPG level < 100 mg/dL), normal-high FPG (FPG level of 100-109 mg/dL), impaired fasting glucose (FPG level of 110-125 mg/dL), and diabetes mellitus (FPG level ≥ 126 mg/dL).
Fasting Plasma Glucose Category and Colorectal Cancer
During a mean follow-up of 1137 ± 824 days, 5566 CRC events occurred. Kaplan-Meier curves and the log-rank test demonstrated that there was a significant difference in the incidence of CRC among the FPG categories (Fig. 2). The incidence rate for CRC events were the lowest in the normal FPG group [10.5 (95% CI 10.1-10.8) per 10 000 person-years], followed by the normal-high FPG group [17.3 (95% CI 16.3-18.3) per 10 000 person-years], IFG group [23.8 (95% CI 21.9-25.9) per 10 000 person-years], and DM group [26.5 (95% CI 23.4-30.0) per 10 000 person-years] (Table 2). Univariate Cox regression analyses showed that the presence of normal-high FPG, IFG, or DM was associated with a higher incidence rate of CRC events than that of normal FPG. Age- and sex-adjusted Cox regression analyses showed that normal-high FPG, IFG, or DM vs normal FPG was associated with a higher incidence rate of CRC events. Multivariable Cox regression analyses showed that the presence of normal-high FPG (HR 1.10, 95% CI 1.03-1.18), IFG (HR 1.24, 95% CI 1.13-1.37), and DM (HR 1.36, 95% CI 1.19-1.55) was associated with a higher incidence rate of CRC than that of normal FPG. The RRR for CRC associated with normal-high FPG, IFG, and DM was 9.4% (95% CI 2.8-15.6), 19.6% (95% CI 11.7-26.9), and 26.4% (95% CI 16.0-35.5).
Frequency of events, corresponding incidence rates, and hazard ratios for colorectal cancer events
| Normal FPG (n = 1 125 647) | Normal-High FPG (n = 210 365) | Impaired Fasting Glucose (n = 74 836) | Diabetes Mellitus (n = 30 463) | P for trend | |
|---|---|---|---|---|---|
| Events, n | 3690 | 1094 | 534 | 248 | — |
| Incidence rate | 10.5 (10.1-10.8) | 17.3 (16.3-18.3) | 23.8 (21.9-25.9) | 26.5 (23.4-30.0) | — |
| Model 1 (unadjusted) | 1 [Reference] | 1.65 (1.54-1.76) | 2.26 (2.07-2.48) | 2.52 (2.22-2.87) | <0.001 |
| Model 2 | 1 [Reference] | 1.14 (1.07-1.22) | 1.32 (1.20-1.45) | 1.46 (1.28-1.66) | <0.001 |
| Model 3 | 1 [Reference] | 1.10 (1.03-1.18) | 1.24 (1.13-1.37) | 1.36 (1.19-1.55) | <0.001 |
| Normal FPG (n = 1 125 647) | Normal-High FPG (n = 210 365) | Impaired Fasting Glucose (n = 74 836) | Diabetes Mellitus (n = 30 463) | P for trend | |
|---|---|---|---|---|---|
| Events, n | 3690 | 1094 | 534 | 248 | — |
| Incidence rate | 10.5 (10.1-10.8) | 17.3 (16.3-18.3) | 23.8 (21.9-25.9) | 26.5 (23.4-30.0) | — |
| Model 1 (unadjusted) | 1 [Reference] | 1.65 (1.54-1.76) | 2.26 (2.07-2.48) | 2.52 (2.22-2.87) | <0.001 |
| Model 2 | 1 [Reference] | 1.14 (1.07-1.22) | 1.32 (1.20-1.45) | 1.46 (1.28-1.66) | <0.001 |
| Model 3 | 1 [Reference] | 1.10 (1.03-1.18) | 1.24 (1.13-1.37) | 1.36 (1.19-1.55) | <0.001 |
The incidence rate was per 10 000 person-years. Unadjusted and adjusted hazard ratios (95% CIs) associated with FPG category are shown. Model 1 is unadjusted. Model 2 includes adjustment for age and sex. Model 3 includes adjustment for age, sex, obesity, high waist circumference, hypertension, dyslipidemia, cigarette smoking, alcohol consumption, and physical inactivity.
Abbreviation: FPG, fasting plasma glucose.
Frequency of events, corresponding incidence rates, and hazard ratios for colorectal cancer events
| Normal FPG (n = 1 125 647) | Normal-High FPG (n = 210 365) | Impaired Fasting Glucose (n = 74 836) | Diabetes Mellitus (n = 30 463) | P for trend | |
|---|---|---|---|---|---|
| Events, n | 3690 | 1094 | 534 | 248 | — |
| Incidence rate | 10.5 (10.1-10.8) | 17.3 (16.3-18.3) | 23.8 (21.9-25.9) | 26.5 (23.4-30.0) | — |
| Model 1 (unadjusted) | 1 [Reference] | 1.65 (1.54-1.76) | 2.26 (2.07-2.48) | 2.52 (2.22-2.87) | <0.001 |
| Model 2 | 1 [Reference] | 1.14 (1.07-1.22) | 1.32 (1.20-1.45) | 1.46 (1.28-1.66) | <0.001 |
| Model 3 | 1 [Reference] | 1.10 (1.03-1.18) | 1.24 (1.13-1.37) | 1.36 (1.19-1.55) | <0.001 |
| Normal FPG (n = 1 125 647) | Normal-High FPG (n = 210 365) | Impaired Fasting Glucose (n = 74 836) | Diabetes Mellitus (n = 30 463) | P for trend | |
|---|---|---|---|---|---|
| Events, n | 3690 | 1094 | 534 | 248 | — |
| Incidence rate | 10.5 (10.1-10.8) | 17.3 (16.3-18.3) | 23.8 (21.9-25.9) | 26.5 (23.4-30.0) | — |
| Model 1 (unadjusted) | 1 [Reference] | 1.65 (1.54-1.76) | 2.26 (2.07-2.48) | 2.52 (2.22-2.87) | <0.001 |
| Model 2 | 1 [Reference] | 1.14 (1.07-1.22) | 1.32 (1.20-1.45) | 1.46 (1.28-1.66) | <0.001 |
| Model 3 | 1 [Reference] | 1.10 (1.03-1.18) | 1.24 (1.13-1.37) | 1.36 (1.19-1.55) | <0.001 |
The incidence rate was per 10 000 person-years. Unadjusted and adjusted hazard ratios (95% CIs) associated with FPG category are shown. Model 1 is unadjusted. Model 2 includes adjustment for age and sex. Model 3 includes adjustment for age, sex, obesity, high waist circumference, hypertension, dyslipidemia, cigarette smoking, alcohol consumption, and physical inactivity.
Abbreviation: FPG, fasting plasma glucose.
Kaplan-Meier curves for colorectal cancer. The cumulative probability of colorectal cancer events for each fasting plasma glucose group was calculated using the Kaplan-Meier method. Log-rank test was used to calculate P-value, and the value was less than 0.001. Participants were categorized as having normal fasting plasma glucose (<100 mg/dL), normal-high fasting plasma glucose (100-109 mg/dL), impaired fasting glucose (110-125 mg/dL), and diabetic mellitus (≥126 mg/dL).
Sensitivity Analyses
First, Figure 3 shows the dose-response relationship between FPG and the risk of incident CRC. The association between FPG and the incidence rate of CRC was modeled through multivariable-adjusted spline regression models with reference point set at FPG of 100 mg/dL. A linear dose-response relationship was observed between FPG and the incidence rate of CRC. FPG per standard deviation (14.7 mg/dL) was associated with incident CRC in the univariate model (HR 1.17, 95% CI 1.15-1.18) and age-sex-adjusted model (HR 1.08, 95% CI 1.06-1.10). The multivariable Cox regression model also showed that FPG per standard deviation was associated with a higher incidence rate of CRC (HR 1.07, 95% CI 1.05-1.09). Second, we conducted a multivariable Cox regression analysis after multiple imputations for missing values. Multivariable Cox regression analysis after multiple imputation included 1 877 713 participants and 6980 CRC events during a mean follow-up period of 1165 ± 845 days and showed that having normal-high FPG (HR 1.12, 95% CI 1.05-1.19), IFG (HR 1.27, 95% CI 1.17-1.38), and DM (HR 1.39, 95% CI 1.24-1.57) was associated with a higher incidence rate of CRC than having normal FPG (Table 3). Third, we excluded participants with a follow-up period for CRC < 365 days (as an induction period), and we analyzed 1 154 621 participants with a follow-up period for CRC ≥ 365 days. Among them, 3817 participants experienced CRC events. The presence of normal-high FPG (HR 1.10, 95% CI 1.01-1.19), IFG (HR 1.22, 95% CI 1.08-1.36), and DM (HR 1.25, 95% CI 1.06-1.47) was associated with a higher incidence rate of CRC than that of normal FPG (Table 4). Fourth, we excluded 335 275 obese participants and analyzed data of 1 106 036 participants. During a mean follow-up of 1140 ± 828 days, 4047 CRC events occurred. The presence of normal-high FPG (HR 1.13, 95% CI 1.04-1.22), IFG (HR 1.16, 95% CI 1.02-1.31), and DM (HR 1.33, 95% CI 1.11-1.59) was associated with a higher incidence rate of CRC than that of normal FPG (Table 5). Fifth, we analyzed the association of FPG category with CRC stratified by sex. The presence of normal-high FPG (HR 1.13, 95% CI 1.04-1.22), IFG (HR 1.28, 95% CI 1.15-1.32), and DM (HR 1.35, 95% CI 1.17-1.56) was associated with a higher incidence rate of CRC than that of normal FPG in men. However, normal-high FPG and IFG were not associated with a higher incidence rate of CRC than that with normal FPG in women. DM was associated with an incidence rate of CRC in women (HR 1.41, 95% CI 1.02-1.96) (Table 6). However, P for interaction was 0.903, suggesting that the association of FPG category and incident CRC was not modified by sex. Sixth, among 5566 participants diagnosed with CRC, we confirmed that 3984 participants (71.6%) underwent surgical treatment or chemotherapy for CRC. Accordingly, we excluded 1582 participants diagnosed with CRC but no treatment history confirmed from the study population. On this population, compared with normal FPG, normal-high FPG (HR 1.10, 95% CI 1.01-1.19), IFG (HR 1.25, 95% CI 1.12-1.40), and DM (HR 1.38, 95% CI 1.18-1.60) were associated with a greater incidence rate of CRC. Seventh, we created a directed acyclic graph using a browser-based software DAGitty (http://www.dagitty.net) (Fig. 4). According to a directed acyclic graph, we excluded hypertension from the multivariable model. Even in this model, normal-high FPG (HR 1.11, 95% CI 1.04-1.19), IFG (HR 1.27, 95% CI 1.15-1.39), and DM (HR 1.39, 95% CI 1.22-1.58) were associated with a greater incidence rate of CRC compared normal FPG.
Frequency of events, corresponding incidence rates, and hazard ratios for colorectal cancer events in participants after multiple imputation
| Normal FPG (n = 1 466 534) | Normal-high FPG (n = 273 996) | Impaired fasting glucose (n = 97 457) | Diabetes mellitus (n = 39 726) | P for trend | |
|---|---|---|---|---|---|
| Events, n | 4617 | 1381 | 670 | 312 | — |
| Incidence rate | 9.8 (95-10.1) | 16.5 (15.6-17.3) | 22.7 (21.0-24.5) | 25.5 (22.8-28.5) | — |
| Model 1 (unadjusted) | 1 [Reference] | 1.67 (1.58-1.78) | 2.30 (2.12-2.50) | 2.59 (2.31-2.91) | <0.001 |
| Model 2 | 1 [Reference] | 1.16 (1.09-1.23) | 1.34 (1.23-1.68) | 1.49 (1.33-1.68) | <0.001 |
| Model 3 | 1 [Reference] | 1.12 (1.05-1.19) | 1.27 (1.17-1.38) | 1.39 (1.24-1.57) | <0.001 |
| Normal FPG (n = 1 466 534) | Normal-high FPG (n = 273 996) | Impaired fasting glucose (n = 97 457) | Diabetes mellitus (n = 39 726) | P for trend | |
|---|---|---|---|---|---|
| Events, n | 4617 | 1381 | 670 | 312 | — |
| Incidence rate | 9.8 (95-10.1) | 16.5 (15.6-17.3) | 22.7 (21.0-24.5) | 25.5 (22.8-28.5) | — |
| Model 1 (unadjusted) | 1 [Reference] | 1.67 (1.58-1.78) | 2.30 (2.12-2.50) | 2.59 (2.31-2.91) | <0.001 |
| Model 2 | 1 [Reference] | 1.16 (1.09-1.23) | 1.34 (1.23-1.68) | 1.49 (1.33-1.68) | <0.001 |
| Model 3 | 1 [Reference] | 1.12 (1.05-1.19) | 1.27 (1.17-1.38) | 1.39 (1.24-1.57) | <0.001 |
We analyzed 1 877 713 participants after multiple imputation for missing values. The incidence rate was per 10 000 person-years. Unadjusted and adjusted hazard ratios (95% CIs) associated with FPG category are shown. Model 1 is unadjusted. Model 2 includes adjustment for age and sex. Model 3 includes adjustment for age, sex, obesity, high waist circumference, hypertension, dyslipidemia, cigarette smoking, alcohol consumption, and physical inactivity.
Abbreviation: FPG, fasting plasma glucose.
Frequency of events, corresponding incidence rates, and hazard ratios for colorectal cancer events in participants after multiple imputation
| Normal FPG (n = 1 466 534) | Normal-high FPG (n = 273 996) | Impaired fasting glucose (n = 97 457) | Diabetes mellitus (n = 39 726) | P for trend | |
|---|---|---|---|---|---|
| Events, n | 4617 | 1381 | 670 | 312 | — |
| Incidence rate | 9.8 (95-10.1) | 16.5 (15.6-17.3) | 22.7 (21.0-24.5) | 25.5 (22.8-28.5) | — |
| Model 1 (unadjusted) | 1 [Reference] | 1.67 (1.58-1.78) | 2.30 (2.12-2.50) | 2.59 (2.31-2.91) | <0.001 |
| Model 2 | 1 [Reference] | 1.16 (1.09-1.23) | 1.34 (1.23-1.68) | 1.49 (1.33-1.68) | <0.001 |
| Model 3 | 1 [Reference] | 1.12 (1.05-1.19) | 1.27 (1.17-1.38) | 1.39 (1.24-1.57) | <0.001 |
| Normal FPG (n = 1 466 534) | Normal-high FPG (n = 273 996) | Impaired fasting glucose (n = 97 457) | Diabetes mellitus (n = 39 726) | P for trend | |
|---|---|---|---|---|---|
| Events, n | 4617 | 1381 | 670 | 312 | — |
| Incidence rate | 9.8 (95-10.1) | 16.5 (15.6-17.3) | 22.7 (21.0-24.5) | 25.5 (22.8-28.5) | — |
| Model 1 (unadjusted) | 1 [Reference] | 1.67 (1.58-1.78) | 2.30 (2.12-2.50) | 2.59 (2.31-2.91) | <0.001 |
| Model 2 | 1 [Reference] | 1.16 (1.09-1.23) | 1.34 (1.23-1.68) | 1.49 (1.33-1.68) | <0.001 |
| Model 3 | 1 [Reference] | 1.12 (1.05-1.19) | 1.27 (1.17-1.38) | 1.39 (1.24-1.57) | <0.001 |
We analyzed 1 877 713 participants after multiple imputation for missing values. The incidence rate was per 10 000 person-years. Unadjusted and adjusted hazard ratios (95% CIs) associated with FPG category are shown. Model 1 is unadjusted. Model 2 includes adjustment for age and sex. Model 3 includes adjustment for age, sex, obesity, high waist circumference, hypertension, dyslipidemia, cigarette smoking, alcohol consumption, and physical inactivity.
Abbreviation: FPG, fasting plasma glucose.
Frequency of events, corresponding incidence rates, and hazard ratios for colorectal cancer events in participants with follow-up ≥365 days
| Normal FPG (n = 901 796) | Normal-high FPG (n = 168 453) | Impaired fasting glucose (n = 60 023) | Diabetes mellitus (n = 24 349) | P for trend | |
|---|---|---|---|---|---|
| Events, n | 2561 | 742 | 356 | 158 | — |
| Incidence rate | 10.2 (9.8-10.6) | 16.8 (15.6-18.0) | 22.8 (20.5-25.3) | 24.0 (20.5-28.0) | — |
| Model 1 (unadjusted) | 1 [Reference] | 1.64 (1.51-1.78) | 2.22 (1.99-2.48) | 2.34 (1.99-2.75) | <0.001 |
| Model 2 | 1 [Reference] | 1.14 (1.04-1.23) | 1.29 (1.15-1.45) | 1.35 (1.14-1.59) | <0.001 |
| Model 3 | 1 [Reference] | 1.10 (1.01-1.19) | 1.22 (1.08-1.36) | 1.25 (1.06-1.47) | <0.001 |
| Normal FPG (n = 901 796) | Normal-high FPG (n = 168 453) | Impaired fasting glucose (n = 60 023) | Diabetes mellitus (n = 24 349) | P for trend | |
|---|---|---|---|---|---|
| Events, n | 2561 | 742 | 356 | 158 | — |
| Incidence rate | 10.2 (9.8-10.6) | 16.8 (15.6-18.0) | 22.8 (20.5-25.3) | 24.0 (20.5-28.0) | — |
| Model 1 (unadjusted) | 1 [Reference] | 1.64 (1.51-1.78) | 2.22 (1.99-2.48) | 2.34 (1.99-2.75) | <0.001 |
| Model 2 | 1 [Reference] | 1.14 (1.04-1.23) | 1.29 (1.15-1.45) | 1.35 (1.14-1.59) | <0.001 |
| Model 3 | 1 [Reference] | 1.10 (1.01-1.19) | 1.22 (1.08-1.36) | 1.25 (1.06-1.47) | <0.001 |
We analyzed 1 154 621 participants with follow-up ≥365 days. The incidence rate was per 10 000 person-years. Unadjusted and adjusted hazard ratios (95% CIs) associated with FPG category are shown. Model 1 is unadjusted. Model 2 includes adjustment for age and sex. Model 3 includes adjustment for age, sex, obesity, high waist circumference, hypertension, dyslipidemia, cigarette smoking, alcohol consumption, and physical inactivity.
Abbreviation: FPG, fasting plasma glucose.
Frequency of events, corresponding incidence rates, and hazard ratios for colorectal cancer events in participants with follow-up ≥365 days
| Normal FPG (n = 901 796) | Normal-high FPG (n = 168 453) | Impaired fasting glucose (n = 60 023) | Diabetes mellitus (n = 24 349) | P for trend | |
|---|---|---|---|---|---|
| Events, n | 2561 | 742 | 356 | 158 | — |
| Incidence rate | 10.2 (9.8-10.6) | 16.8 (15.6-18.0) | 22.8 (20.5-25.3) | 24.0 (20.5-28.0) | — |
| Model 1 (unadjusted) | 1 [Reference] | 1.64 (1.51-1.78) | 2.22 (1.99-2.48) | 2.34 (1.99-2.75) | <0.001 |
| Model 2 | 1 [Reference] | 1.14 (1.04-1.23) | 1.29 (1.15-1.45) | 1.35 (1.14-1.59) | <0.001 |
| Model 3 | 1 [Reference] | 1.10 (1.01-1.19) | 1.22 (1.08-1.36) | 1.25 (1.06-1.47) | <0.001 |
| Normal FPG (n = 901 796) | Normal-high FPG (n = 168 453) | Impaired fasting glucose (n = 60 023) | Diabetes mellitus (n = 24 349) | P for trend | |
|---|---|---|---|---|---|
| Events, n | 2561 | 742 | 356 | 158 | — |
| Incidence rate | 10.2 (9.8-10.6) | 16.8 (15.6-18.0) | 22.8 (20.5-25.3) | 24.0 (20.5-28.0) | — |
| Model 1 (unadjusted) | 1 [Reference] | 1.64 (1.51-1.78) | 2.22 (1.99-2.48) | 2.34 (1.99-2.75) | <0.001 |
| Model 2 | 1 [Reference] | 1.14 (1.04-1.23) | 1.29 (1.15-1.45) | 1.35 (1.14-1.59) | <0.001 |
| Model 3 | 1 [Reference] | 1.10 (1.01-1.19) | 1.22 (1.08-1.36) | 1.25 (1.06-1.47) | <0.001 |
We analyzed 1 154 621 participants with follow-up ≥365 days. The incidence rate was per 10 000 person-years. Unadjusted and adjusted hazard ratios (95% CIs) associated with FPG category are shown. Model 1 is unadjusted. Model 2 includes adjustment for age and sex. Model 3 includes adjustment for age, sex, obesity, high waist circumference, hypertension, dyslipidemia, cigarette smoking, alcohol consumption, and physical inactivity.
Abbreviation: FPG, fasting plasma glucose.
Frequency of events, corresponding incidence rates, and hazard ratios for colorectal cancer events in nonobese participants
| Normal FPG (n = 911 207) | Normal-high FPG (n = 138 831) | Impaired fasting glucose (n = 41 886) | Diabetes mellitus (n = 14 112) | P for trend | |
|---|---|---|---|---|---|
| Events, n | 2894 | 737 | 290 | 126 | — |
| Incidence rate | 10.1 (9.8-10.5) | 17.6 (16.4-18.9) | 22.9 (20.4-25.7) | 28.6 (24.0-34.0) | — |
| Model 1 (unadjusted) | 1 [Reference] | 1.73 (1.60-1.88) | 2.25 (2.00-2.54) | 2.82 (2.36-3.37) | <0.001 |
| Model 2 | 1 [Reference] | 1.15 (1.06-1.25) | 1.21 (1.07-1.37) | 1.40 (1.17-1.68) | <0.001 |
| Model 3 | 1 [Reference] | 1.13 (1.04-1.22) | 1.16 (1.02-1.31) | 1.33 (1.11-1.59) | <0.001 |
| Normal FPG (n = 911 207) | Normal-high FPG (n = 138 831) | Impaired fasting glucose (n = 41 886) | Diabetes mellitus (n = 14 112) | P for trend | |
|---|---|---|---|---|---|
| Events, n | 2894 | 737 | 290 | 126 | — |
| Incidence rate | 10.1 (9.8-10.5) | 17.6 (16.4-18.9) | 22.9 (20.4-25.7) | 28.6 (24.0-34.0) | — |
| Model 1 (unadjusted) | 1 [Reference] | 1.73 (1.60-1.88) | 2.25 (2.00-2.54) | 2.82 (2.36-3.37) | <0.001 |
| Model 2 | 1 [Reference] | 1.15 (1.06-1.25) | 1.21 (1.07-1.37) | 1.40 (1.17-1.68) | <0.001 |
| Model 3 | 1 [Reference] | 1.13 (1.04-1.22) | 1.16 (1.02-1.31) | 1.33 (1.11-1.59) | <0.001 |
We analyzed 1 106 036 nonobese participants. The incidence rate was per 10 000 person-years. Unadjusted and adjusted hazard ratios (95% CIs) associated with FPG category are shown. Model 1 is unadjusted. Model 2 includes adjustment for age and sex. Model 3 includes adjustment for age, sex, high waist circumference, hypertension, dyslipidemia, cigarette smoking, alcohol consumption, and physical inactivity.
Abbreviation: FPG, fasting plasma glucose.
Frequency of events, corresponding incidence rates, and hazard ratios for colorectal cancer events in nonobese participants
| Normal FPG (n = 911 207) | Normal-high FPG (n = 138 831) | Impaired fasting glucose (n = 41 886) | Diabetes mellitus (n = 14 112) | P for trend | |
|---|---|---|---|---|---|
| Events, n | 2894 | 737 | 290 | 126 | — |
| Incidence rate | 10.1 (9.8-10.5) | 17.6 (16.4-18.9) | 22.9 (20.4-25.7) | 28.6 (24.0-34.0) | — |
| Model 1 (unadjusted) | 1 [Reference] | 1.73 (1.60-1.88) | 2.25 (2.00-2.54) | 2.82 (2.36-3.37) | <0.001 |
| Model 2 | 1 [Reference] | 1.15 (1.06-1.25) | 1.21 (1.07-1.37) | 1.40 (1.17-1.68) | <0.001 |
| Model 3 | 1 [Reference] | 1.13 (1.04-1.22) | 1.16 (1.02-1.31) | 1.33 (1.11-1.59) | <0.001 |
| Normal FPG (n = 911 207) | Normal-high FPG (n = 138 831) | Impaired fasting glucose (n = 41 886) | Diabetes mellitus (n = 14 112) | P for trend | |
|---|---|---|---|---|---|
| Events, n | 2894 | 737 | 290 | 126 | — |
| Incidence rate | 10.1 (9.8-10.5) | 17.6 (16.4-18.9) | 22.9 (20.4-25.7) | 28.6 (24.0-34.0) | — |
| Model 1 (unadjusted) | 1 [Reference] | 1.73 (1.60-1.88) | 2.25 (2.00-2.54) | 2.82 (2.36-3.37) | <0.001 |
| Model 2 | 1 [Reference] | 1.15 (1.06-1.25) | 1.21 (1.07-1.37) | 1.40 (1.17-1.68) | <0.001 |
| Model 3 | 1 [Reference] | 1.13 (1.04-1.22) | 1.16 (1.02-1.31) | 1.33 (1.11-1.59) | <0.001 |
We analyzed 1 106 036 nonobese participants. The incidence rate was per 10 000 person-years. Unadjusted and adjusted hazard ratios (95% CIs) associated with FPG category are shown. Model 1 is unadjusted. Model 2 includes adjustment for age and sex. Model 3 includes adjustment for age, sex, high waist circumference, hypertension, dyslipidemia, cigarette smoking, alcohol consumption, and physical inactivity.
Abbreviation: FPG, fasting plasma glucose.
Frequency of events, corresponding incidence rates, and hazard ratios for colorectal cancer events
| Normal FPG | Normal-high FPG | Impaired fasting glucose | Diabetes mellitus | P for trend | |
|---|---|---|---|---|---|
| Men | |||||
| Participants, n | 583 344 | 149 754 | 57 640 | 24 474 | |
| Events, n | 2189 | 850 | 449 | 210 | — |
| Incidence rate | 11.6 (11.1-12.1) | 18.2 (17.1-19.5) | 25.3 (23.1-27.7) | 27.3 (23.9-31.3) | — |
| Model 1 (unadjusted) | 1 [Reference] | 1.58 (1.46-1.71) | 2.18 (1.97-2.42) | 2.36 (2.05-2.72) | <0.001 |
| Model 2 | 1 [Reference] | 1.17 (1.08-1.27) | 1.37 (1.23-1.52) | 1.46 (1.26-1.68) | <0.001 |
| Model 3 | 1 [Reference] | 1.13 (1.04-1.22) | 1.28 (1.15-1.42) | 1.35 (1.17-1.56) | <0.001 |
| Women | |||||
| Participants, n | 542 303 | 60 611 | 17 196 | 5989 | |
| Events, n | 1501 | 244 | 85 | 38 | — |
| Incidence rate | 9.2 (8.7-9.7) | 14.6 (12.9-16.6) | 18.1 (14.7-22.4) | 22.8 (16.6-31.4) | — |
| Model 1 (unadjusted) | 1 [Reference] | 1.57 (1.37-1.80) | 1.94 (1.56-2.42) | 2.46 (1.78-3.39) | <0.001 |
| Model 2 | 1 [Reference] | 1.07 (0.93-1.22) | 1.12 (0.90-1.40) | 1.47 (1.06-2.03) | 0.021 |
| Model 3 | 1 [Reference] | 1.05 (0.91-1.20) | 1.09 (0.87-1.36) | 1.41 (1.02-1.96) | 0.067 |
| Normal FPG | Normal-high FPG | Impaired fasting glucose | Diabetes mellitus | P for trend | |
|---|---|---|---|---|---|
| Men | |||||
| Participants, n | 583 344 | 149 754 | 57 640 | 24 474 | |
| Events, n | 2189 | 850 | 449 | 210 | — |
| Incidence rate | 11.6 (11.1-12.1) | 18.2 (17.1-19.5) | 25.3 (23.1-27.7) | 27.3 (23.9-31.3) | — |
| Model 1 (unadjusted) | 1 [Reference] | 1.58 (1.46-1.71) | 2.18 (1.97-2.42) | 2.36 (2.05-2.72) | <0.001 |
| Model 2 | 1 [Reference] | 1.17 (1.08-1.27) | 1.37 (1.23-1.52) | 1.46 (1.26-1.68) | <0.001 |
| Model 3 | 1 [Reference] | 1.13 (1.04-1.22) | 1.28 (1.15-1.42) | 1.35 (1.17-1.56) | <0.001 |
| Women | |||||
| Participants, n | 542 303 | 60 611 | 17 196 | 5989 | |
| Events, n | 1501 | 244 | 85 | 38 | — |
| Incidence rate | 9.2 (8.7-9.7) | 14.6 (12.9-16.6) | 18.1 (14.7-22.4) | 22.8 (16.6-31.4) | — |
| Model 1 (unadjusted) | 1 [Reference] | 1.57 (1.37-1.80) | 1.94 (1.56-2.42) | 2.46 (1.78-3.39) | <0.001 |
| Model 2 | 1 [Reference] | 1.07 (0.93-1.22) | 1.12 (0.90-1.40) | 1.47 (1.06-2.03) | 0.021 |
| Model 3 | 1 [Reference] | 1.05 (0.91-1.20) | 1.09 (0.87-1.36) | 1.41 (1.02-1.96) | 0.067 |
The incidence rate was per 10 000 person-years. Unadjusted and adjusted hazard ratios (95% CIs) associated with FPG category are shown. Model 1 is unadjusted. Model 2 includes adjustment for age. Model 3 includes adjustment for age, obesity, high waist circumference, hypertension, dyslipidemia, cigarette smoking, alcohol consumption, and physical inactivity.
Abbreviation: FPG, fasting plasma glucose.
Frequency of events, corresponding incidence rates, and hazard ratios for colorectal cancer events
| Normal FPG | Normal-high FPG | Impaired fasting glucose | Diabetes mellitus | P for trend | |
|---|---|---|---|---|---|
| Men | |||||
| Participants, n | 583 344 | 149 754 | 57 640 | 24 474 | |
| Events, n | 2189 | 850 | 449 | 210 | — |
| Incidence rate | 11.6 (11.1-12.1) | 18.2 (17.1-19.5) | 25.3 (23.1-27.7) | 27.3 (23.9-31.3) | — |
| Model 1 (unadjusted) | 1 [Reference] | 1.58 (1.46-1.71) | 2.18 (1.97-2.42) | 2.36 (2.05-2.72) | <0.001 |
| Model 2 | 1 [Reference] | 1.17 (1.08-1.27) | 1.37 (1.23-1.52) | 1.46 (1.26-1.68) | <0.001 |
| Model 3 | 1 [Reference] | 1.13 (1.04-1.22) | 1.28 (1.15-1.42) | 1.35 (1.17-1.56) | <0.001 |
| Women | |||||
| Participants, n | 542 303 | 60 611 | 17 196 | 5989 | |
| Events, n | 1501 | 244 | 85 | 38 | — |
| Incidence rate | 9.2 (8.7-9.7) | 14.6 (12.9-16.6) | 18.1 (14.7-22.4) | 22.8 (16.6-31.4) | — |
| Model 1 (unadjusted) | 1 [Reference] | 1.57 (1.37-1.80) | 1.94 (1.56-2.42) | 2.46 (1.78-3.39) | <0.001 |
| Model 2 | 1 [Reference] | 1.07 (0.93-1.22) | 1.12 (0.90-1.40) | 1.47 (1.06-2.03) | 0.021 |
| Model 3 | 1 [Reference] | 1.05 (0.91-1.20) | 1.09 (0.87-1.36) | 1.41 (1.02-1.96) | 0.067 |
| Normal FPG | Normal-high FPG | Impaired fasting glucose | Diabetes mellitus | P for trend | |
|---|---|---|---|---|---|
| Men | |||||
| Participants, n | 583 344 | 149 754 | 57 640 | 24 474 | |
| Events, n | 2189 | 850 | 449 | 210 | — |
| Incidence rate | 11.6 (11.1-12.1) | 18.2 (17.1-19.5) | 25.3 (23.1-27.7) | 27.3 (23.9-31.3) | — |
| Model 1 (unadjusted) | 1 [Reference] | 1.58 (1.46-1.71) | 2.18 (1.97-2.42) | 2.36 (2.05-2.72) | <0.001 |
| Model 2 | 1 [Reference] | 1.17 (1.08-1.27) | 1.37 (1.23-1.52) | 1.46 (1.26-1.68) | <0.001 |
| Model 3 | 1 [Reference] | 1.13 (1.04-1.22) | 1.28 (1.15-1.42) | 1.35 (1.17-1.56) | <0.001 |
| Women | |||||
| Participants, n | 542 303 | 60 611 | 17 196 | 5989 | |
| Events, n | 1501 | 244 | 85 | 38 | — |
| Incidence rate | 9.2 (8.7-9.7) | 14.6 (12.9-16.6) | 18.1 (14.7-22.4) | 22.8 (16.6-31.4) | — |
| Model 1 (unadjusted) | 1 [Reference] | 1.57 (1.37-1.80) | 1.94 (1.56-2.42) | 2.46 (1.78-3.39) | <0.001 |
| Model 2 | 1 [Reference] | 1.07 (0.93-1.22) | 1.12 (0.90-1.40) | 1.47 (1.06-2.03) | 0.021 |
| Model 3 | 1 [Reference] | 1.05 (0.91-1.20) | 1.09 (0.87-1.36) | 1.41 (1.02-1.96) | 0.067 |
The incidence rate was per 10 000 person-years. Unadjusted and adjusted hazard ratios (95% CIs) associated with FPG category are shown. Model 1 is unadjusted. Model 2 includes adjustment for age. Model 3 includes adjustment for age, obesity, high waist circumference, hypertension, dyslipidemia, cigarette smoking, alcohol consumption, and physical inactivity.
Abbreviation: FPG, fasting plasma glucose.
Restricted cubic spline. Restricted cubic spline showed the linear dose-response relationship between fasting plasma glucose and the risk of colorectal cancer. The reference point was set at fasting plasma glucose of 100 mg/dL. Hazard ratios were adjusted for age, sex, obesity, high waist circumference, hypertension, dyslipidemia, cigarette smoking, alcohol consumption, and physical inactivity.
Directed acyclic graph. Directed acyclic graph model of incident colorectal cancer. Abbreviation: CRC, colorectal cancer.
Discussion
In this nationwide health claims database that enrolled adults who had annual health check-ups, normal-high FPG, IFG, and DM were associated with a higher incidence rate of CRC events. This association was observed among adults who did not have obesity. The risk of CRC started to increase from normal-high FPG, and further increased in IFG and DM in men, whereas the risk of CRC increased only in women with DM. Restricted cubic spline showed a linear dose-dependent association between FPG and incident CRC.
The relationship between the presence of DM or higher FPG and incident CRC has been widely investigated (3-6). For example, analysis of the Japan Public Health Center-based Prospective Study, including 97 771 general Japanese individuals demonstrated that DM was associated with incident colon cancer (HR 1.36, 95% CI 1.00–1.85) in men, but not in women (29). A multiethnic population-based prospective cohort in the United States, including 3459 incident CRC cases showed that DM was associated with a greater risk of CRC (relative risk 1.19, 95% CI 1.09-1.29) (34). The Korean Cancer Prevention Study II showed that higher FPG was associated with an elevated risk of CRC in men (HR 1.51, 95% CI 1.11-2.05) (30). However, these previous studies did not exclude subjects with a prior history of CRC, colorectal polyps, or inflammatory bowel disease. In addition, because they potentially included subjects taking antidiabetic medications, FPG could have been influenced by these medications. Furthermore, the potential proportion of CRC that can be prevented by normalizing hyperglycemia has not been estimated yet.
The current study extends our knowledge by demonstrating that in individuals not taking antidiabetic medication and those who do not have colorectal polyps, Crohn’s disease, and/or ulcerative colitis, normal-high FPG and IFG, and DM were associated with a greater incidence rate of CRC events. FPG as a continuous value was also associated with incident CRC, and the restricted cubic spline curve presented a linear relationship between FPG and the incidence rate of CRC. Normal-high FPG and IFG were significantly associated with a higher incidence rate of CRC events in men but not in women. The risk of CRC in women is elevated only in patients with DM. However, given the value of P for interaction, the association of FPG category with an incidence rate of CRC was not modified by sex, and the difference in the result between men and women would be due to the difference in the incidence rate of CRC between men and women.
Although several pathological mechanisms underlying the association between hyperglycemia and CRC, such as increased DNA damage, interaction with the cell cycle, and promotion of migration and invasion of cancer cells (35-39), could be suggested, the findings of this observational study are unable to conclude a causal relationship between hyperglycemia and CRC. Detailed pathological mechanisms should be clarified in future studies.
The next issue to be solved is whether lowering or normalizing FPG could reduce the incidence rate of CRC. For this purpose, a well-designed prospective study or randomized clinical trial is required to explore potential efficacies of nonpharmacological intervention or pharmacological intervention for the prevention of CRC. However, such investigations would be practically not feasible due to ethical concerns. Thus, careful analyses using real-world available data would be an alternative approach to provide clinical evidence on whether FPG-lowering treatment could lower the incidence rate of CRC.
Our results have several clinical implications. Given the high prevalence and mortality of CRC, early detection and appropriate therapeutic interventions are indispensable for CRC. From this point of view, the optimal risk stratification in the general population is the first step. Our results suggest that measuring FPG could identify individuals at a high risk of subsequent CRC. Further studies are required to assess whether fecal occult blood testing or colonoscopy could be effective for early CRC detection particularly in individuals with hyperglycemia and DM.
We acknowledge several limitations to be addressed. Although we performed multivariable Cox regression analysis, unmeasured confounders, possible residual bias, and time-varying covariates could influence our results. For example, eating habits would be associated with an incidence of CRC. However, we were unable to assess eating habits in this study. Our sensitivity analysis excluding individuals with a follow-up period shorter than 1 year (as an induction period) confirmed our main results. However, the observational period was relatively short, and therefore, the possibility of a latent CRC could not be eliminated. Although we confirmed our results after multiple imputation for missing data, the substantial proportion of missing data should be considered as a major study limitation. Since the JMDC Claims Database mainly included an employed working-age population, we should consider selection bias (eg, healthy worker bias). Similarly, because the population in this study was young, the incidence rate of CRC is relatively lower than that found in other epidemiological data from Japanese general population (40). Further investigations are warranted to generalize the results of the present study to other populations. In the JMDC Claims Database, CRC was identified using ICD-10 code. Therefore, there remains uncertainty regarding the accuracy of the diagnoses for CRC. Detection bias may cause overestimation of true association between FPG and CRC events. For example, participants with hyperglycemia would be more likely to receive medical care and have CRC detected more frequently than those with normal FPG. Thus, the results of RRR should be carefully interpreted.
In conclusion, normal-high FPG, IFG, and DM were associated with a higher risk of incident CRC events. In men, the risk of CRC started to increase from normal-high FPG, whereas the risk of CRC increased only in women with DM. Detecting hyperglycemia can contribute to identifying individuals who are at a higher risk of subsequent CRC events.
Acknowledgments
Funding: This work was supported by grants from the Ministry of Health, Labour and Welfare, Japan (19AA2007 and H30-Policy-Designated-004) and the Ministry of Education, Culture, Sports, Science and Technology, Japan (17H04141). H.K. and K.F. received research funding and scholarship funds from Medtronic Japan Co., LTD; Biotronik Japan; SIMPLEX QUANTUM CO., LTD; Boston Scientific Japan Co., LTD; and Fukuda Denshi, Central Tokyo Co., LTD.
Author Contributions:
H.K., Y.Y., and H.M. conceived the study design. H.I., K.M., H.S., H.K., and T.K. analyzed data. K.F., N.M., T.J., and N.T. reviewed the manuscript. A.N., K.N., and H.Y. supervised the study. I.K. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Additional Information
Disclosures: Research funding and scholarship funds (H.K. and K.F.) from Medtronic Japan Co., LTD, Abbott Medical Japan Co., LTD, Boston Scientific Japan Co., LTD, and Fukuda Denshi, Central Tokyo Co., LTD. Other authors have nothing to disclose. No potential conflicts of interest relevant to this study were reported.
Data Availability
Some or all data sets generated during and/or analyzed during the current study are not publicly available. This database is available for anyone who purchases it from the JMDC inc (https://www.jmdc.co.jp/en/index).
