Diabetes Mellitus and Risk of Colorectal Cancer: A Meta-Analysis

Abstract

Background: Diabetes has been associated with an increased risk of colorectal cancer in most, but not all, studies. Findings have also been inconclusive with regard to sex and subsite in the colorectum. To resolve these inconsistencies, we conducted a meta-analysis of published data on the association between diabetes and the incidence and mortality of colorectal cancer. Methods: We identified studies by a literature search of Medline from January 1, 1966, through July 31, 2005, and by searching the reference lists of pertinent articles. Summary relative risks (RRs) with 95% confidence intervals (CIs) were calculated with a random-effects model. All statistical tests were two-sided. Results: Analysis of 15 studies (six case–control and nine cohort studies), including 2 593 935 participants, found that diabetes was associated with an increased risk of colorectal cancer, compared with no diabetes (summary RR of colorectal cancer incidence = 1.30, 95% CI = 1.20 to 1.40), without heterogeneity between studies ( P_{heterogeneity} = .21). These results were consistent between case–control and cohort studies and between studies conducted in the United States and in Europe. The association between diabetes and colorectal cancer incidence did not differ statistically significantly by sex (summary RR among women = 1.33, 95% CI = 1.23 to 1.44; summary RR among men = 1.29, 95% CI = 1.15 to 1.44; P_{heterogeneity} = .26) or by cancer subsite (summary RR for colon = 1.43, 95% CI = 1.28 to 1.60; summary RR for rectum = 1.33, 95% CI = 1.14 to 1.54; P_{heterogeneity} = .42). Diabetes was positively associated with colorectal cancer mortality (summary RR = 1.26, 95% CI = 1.05 to 1.50), but there was evidence for heterogeneity between studies ( P_{heterogeneity} = .04). Conclusions: Our findings strongly support a relationship between diabetes and increased risk of colon and rectal cancer in both women and men.

Type 2 diabetes mellitus and colorectal cancer are major causes of morbidity and mortality in the United States and other Western countries ( 1 – 3 ) . Dietary and lifestyle risk factors for developing insulin resistance and type 2 diabetes, such as Western diet, physical inactivity, and obesity, have also been linked to an increased risk of colon cancer ( 4 – 7 ) . On the basis of the overlapping risk factors, it has been hypothesized that diabetes itself may be a risk factor for colon cancer ( 8 ) .

Although epidemiologic studies of the relationship of diabetes with the risk of colorectal cancer are not entirely consistent, most studies are compatible with a positive association. In some studies, the association appeared to be stronger for colon cancer than for rectal cancer ( 9 – 12 ) , or the association was observed only in women ( 10 , 13 , 14 ) . The interpretation of these findings, however, has been hampered by the low frequency at which both diseases occur in the same individual, which results in the lack of statistical power to adequately analyze this association.

To provide a quantitative assessment of the association between diabetes and risk of colorectal cancer, we conducted a meta-analysis of case–control and cohort studies. We also evaluated whether the association varied by sex and by cancer subsite (colon versus rectum and proximal colon versus distal colon).

M ATERIALS AND M ETHODS

Search Strategy

We identified studies by a literature search of the Medline database (from January 1, 1966, through July 31, 2005) with the following medical subject heading terms and/or text words: “diabetes mellitus,” “diabetes,” “colorectal cancer,” “colorectal neoplasm,” “colon cancer,” “colon neoplasm,” “rectal cancer,” and “rectal neoplasm.” We also reviewed reference lists of the identified publications for additional pertinent studies. No language restrictions were imposed.

Inclusion and Exclusion Criteria

The 25 studies considered for inclusion in this meta-analysis were case–control and cohort studies on the association between diabetes and the incidence of or mortality from colon, rectal, or colorectal cancer ( 9 – 33 ) . Studies were excluded if they provided only an estimate of effect with no means to calculate a confidence interval or if the estimates were not adjusted by age. When there were multiple publications from the same population or cohort, only data from the most recent report were included. We excluded three candidate studies ( 15 – 17 ) because of overlapping publications and one study ( 18 ) that reported only crude data that were not adjusted by age.

Data Extraction

The data that we extracted included publication data (the first author's last name, year of publication, and country of population studied), study design, number of exposed and unexposed subjects, control source (in case–control studies), follow-up period (for cohort studies), type of diabetes (type 1 or 2), risk estimates with their corresponding confidence intervals, and variables controlled for by matching or in the multivariable model. From each study, we extracted the risk estimates that reflected the greatest degree of control for potential confounders.

Statistical Analysis

We divided epidemiologic studies of the relationship between diabetes and colorectal cancer risk into three general types according to measure of relative risk (RR): case–control studies (odds ratio), cohort studies (incidence and/or mortality rate ratio), and cohort studies with an external comparison group (standardized incidence and/or mortality ratio). We conducted separate meta-analyses of colorectal cancer incidence and mortality. The measure of effect of interest is the relative risk. Because colorectal cancer is rare, the odds ratio in case–control studies and rate ratios in cohort studies yield similar estimates of relative risk ( 34 ) . Cohort studies that reported standardized incidence/mortality ratio were analyzed separately.

Summary relative risk estimates with their corresponding 95% confidence intervals (CIs) were derived with the method of DerSimonian and Laird ( 35 ) by use of the assumptions of a random effects model, which incorporates between-study variability. We calculated a pooled relative risk and its corresponding 95% confidence interval for studies that reported only sex- and/or subsite-specific relative risks. Statistical heterogeneity between studies was evaluated with Cochran's Q test and the I² statistic ( 36 ) . Publication bias was assessed by constructing a funnel plot ( 37 ) , by Begg's adjusted rank correlation test, and by Egger's regression asymmetry test ( 38 ) .

For case–control studies and cohort studies that reported incidence rate ratios, we conducted subgroup meta-analyses to examine potential sources of heterogeneity, including study design, type of control subjects in case–control studies, sex, cancer subsite, and duration of follow-up for cohort studies. Statistical analyses were carried out with Stata, version 8.0 (Stata Corp, College Station, TX). P values that were less than .05 were considered statistically significant. All statistical tests were two-sided.

R ESULTS

Study Characteristics

Twenty-one independent studies met the predefined inclusion criteria. Of these 21 studies, six were case–control studies ( 10 , 13 , 19 , 26 , 27 , 33 ) ( Table 1 ), 11 were cohort studies that used incidence and/or mortality rate ratios as the measure of relative risk ( 9 , 11 , 14 , 20 , 22 – 24 , 28 , 30 – 32 ) ( Table 2 ), and four were cohort studies that used standardized incidence and/or mortality ratio as the measure of relative risk ( 12 , 21 , 25 , 29 ) ( Table 3 ). Eleven studies were conducted in the United States, eight in Europe, one in Australia, and one in Korea. Of the 15 cohort studies, incident colorectal cancer was the outcome in seven, mortality from colon, rectal, or colorectal cancer was the outcome in three, and colorectal cancer incidence and mortality were reported in five. In the primary meta-analysis of diabetes and colorectal cancer incidence, we included all six case–control studies ( 10 , 13 , 19 , 26 , 27 , 33 ) and the nine cohort studies that reported incidence rate ratios ( 9 , 11 , 14 , 20 , 22 , 24 , 28 , 30 , 31 ) . These 15 studies included a total of 2 593 935 participants. The three cohort studies ( 12 , 21 , 25 ) that reported standardized incidence ratio were analyzed separately. For the meta-analysis of diabetes and colorectal cancer mortality, we included the six cohort studies that reported mortality rate ratio ( 11 , 20 , 23 , 28 , 30 , 32 ) . These six studies enrolled a total of 2 523 580 participants.

Colorectal Cancer Incidence

Individual study results and the overall summary results for the six case–control and nine cohort studies of diabetes and colorectal cancer incidence are shown in Fig. 1 . Eight of these 15 studies found a statistically significant positive association between diabetes and colorectal cancer incidence (range of individual RRs = 1.02 to 2.78; summary RR for all 15 studies = 1.30, 95% CI = 1.20 to 1.40). No heterogeneity among studies was found ( Q = 17.9; P_{heterogeneity} = .21; I² = 21.8%). In a sensitivity analysis in which one study at a time was excluded and the rest were analyzed, we detected a statistically significant positive association between diabetes and colorectal cancer incidence (range of summary RRs = 1.25 to 1.36; the lower limit of the 95% CI never crossed 1.0).

We then conducted subgroup meta-analyses by study design, geographical area, sex, cancer subsite, and duration of follow-up ( Table 4 ). The association between diabetes and colorectal cancer incidence was somewhat stronger in case–control studies (summary RR = 1.36; 95% CI = 1.23 to 1.50) than in cohort studies (summary RR = 1.29; 95% CI = 1.16 to 1.43), but there was no heterogeneity among study types ( P_{heterogeneity} = .08). In case–control studies, the source of control subjects did not statistically significantly affect the magnitude of the association. Results were consistent for studies conducted in the United States and in Europe. Results were also consistent for studies in women (summary RR among women = 1.33, 95% CI = 1.23 to 1.44) and in men (summary RR among men = 1.29, 95% CI = 1.15 to 1.44) ( P_{heterogeneity} = .26). For cancer subsites, the summary estimate was similar for colon cancer (RR = 1.43, 95% CI = 1.28 to 1.60) and for rectal cancer (RR = 1.33, 95% CI = 1.14 to 1.54) ( P_{heterogeneity} = .42). There was also no statistically significant difference between subsites in the colon (i.e., proximal colon versus distal colon). Stratifying cohort studies by duration of follow-up resulted in no evidence of heterogeneity. Finally, the summary estimate was similar ( P_{heterogeneity} = .73) for studies (four case–control and four cohort) published before 2000 (RR = 1.28, 95% CI = 1.16 to 1.42) and for studies (two case–control and five cohort) published after 2000 (RR = 1.36, 95% CI = 1.17 to 1.58).

Physical activity and body mass index are potentially the most important known confounders of the positive association between diabetes and colorectal cancer risk. When we restricted the meta-analysis to studies that controlled for these variables ( 10 , 11 , 14 , 20 , 22 , 31 , 33 ) , we found a positive association between diabetes and colorectal cancer (summary RR = 1.34, 95% CI = 1.20 to 1.49), without statistically significant heterogeneity among studies ( Q = 2.05; P_{heterogeneity} = 0.91). Three studies reported both age-adjusted and multivariable-adjusted relative risks ( 11 , 20 , 31 ) . When we restricted the meta-analysis to those studies, the positive association between diabetes and colorectal cancer remained (summary RR adjusted for age only = 1.34, 95% CI = 1.12 to 1.60), and the estimate was identical to estimates that were adjusted for physical activity and body mass index, as well as other potential confounders (RR = 1.34, 95% CI = 1.18 to 1.53), indicating a lack of confounding.

A positive association was observed between diabetes and colorectal cancer incidence in the three cohort studies that reported standardized incidence ratios (summary RR = 1.22, 95% CI = 1.07 to 1.40) ( Table 3 ). The test for heterogeneity among these three studies was not statistically significant ( Q = 4.52; P_{heterogeneity} = .10; I² = 55.8%).

Colorectal Cancer Mortality

Of six cohort studies of diabetes and mortality from colon ( 23 ) or colorectal ( 11 , 20 , 28 , 30 , 32 ) cancer, three ( 11 , 30 , 32 ) reported a statistically significant positive association, and one ( 28 ) observed a nonstatistically significant 3.6-fold (RR = 3.60, 95% CI = 0.81 to 15.89) increase in colorectal cancer risk associated with diabetes ( Table 2 ). When all six studies were analyzed, a positive association between diabetes and mortality from colorectal cancer was found (summary RR = 1.26, 95% CI = 1.05 to 1.50). However, there was statistically significant heterogeneity among studies ( Q = 11.59; P_{heterogeneity} = .04; I² = 56.8%). A sensitivity analysis identified the study by Hu et al. ( 11 ) as contributing most to the heterogeneity. In an analysis excluding this study, the association between diabetes and mortality from colorectal cancer was weaker (summary RR = 1.19, 95% CI = 1.10 to 1.28), and the test for heterogeneity was not statistically significant ( Q = 4.03; P_{heterogeneity} = .40; I² = 0.6%). The association between diabetes and colorectal cancer mortality did not differ statistically significantly by sex ( Q = 0.10; P_{heterogeneity} = .75).

Of two cohort studies that reported standardized mortality ratios ( 21 , 29 ) ( Table 3 ), one ( 21 ) reported a statistically significant approximately 1.5-fold increased risk of death from colorectal cancer among diabetic patients. The other study found no statistically significant association between diabetes and rectal cancer mortality ( 29 ) .

Publication Bias

There was no funnel plot asymmetry for the association between diabetes and colorectal cancer risk (data not shown). P values for Begg's adjusted rank correlation test and Egger's regression asymmetry test were .79 and .28, respectively, indicating a low probability of publication bias.

D ISCUSSION

Findings from this meta-analysis indicate that individuals with diabetes have an approximately 30% increased relative risk of developing colorectal cancer compared with nondiabetic individuals. The results were consistent for case–control and cohort studies and for studies carried out in the United States and in Europe. The association was observed in both women and men and for all subsites in the colorectum.

Our analysis must be interpreted in the context of the limitations of the available data. Most of the studies did not distinguish between type 1 and type 2 diabetes. Because type 1 diabetes [which accounts for 5% to 10% of all diagnosed cases of diabetes ( 1 ) ] may not be related to colorectal cancer ( 12 , 39 ) , the magnitude of the association between diabetes and colorectal cancer risk may have been slightly underestimated. In addition, because diabetes is an underdiagnosed disease, some degree of misclassification of exposure to diabetes is likely to have occurred. Such nondifferential misclassification would also tend to attenuate the true relationship between diabetes and colorectal cancer. As in any meta-analysis, the possibility of publication bias is of concern. However, the results obtained from funnel plot analysis and formal statistical tests did not provide evidence for such bias.

Type 2 diabetes and colorectal cancer share similar risk factors, including physical inactivity and obesity ( 8 ) . Thus, the observed increased risk of colorectal cancer associated with a history of diabetes may reflect confounding by these risk factors. However, a positive association between diabetes and colorectal cancer risk remained when we limited the meta-analysis to studies that controlled for physical activity and body mass index (summary RR = 1.34, 95% CI = 1.20 to 1.49).

Discrepancies among studies investigating the relationship of diabetes with colorectal cancer risk according to sex and cancer subsite may be attributable to small sample sizes that resulted in insufficient statistical power to detect some relationships in the individual studies. Because this meta-analysis included a large number of studies, we could assess the association according to sex and cancer subsite with high precision.

A relationship between diabetes and risk of colorectal cancer is biologically plausible. Type 2 diabetes is characterized by increased insulin concentrations during the early stage of the disease. Hyperinsulinemia ( 8 ) or factors related to insulin resistance, such as hyperglycemia or hypertriglyceridemia ( 40 ) , have been associated with colorectal carcinogenesis. Insulin can stimulate cell proliferation through a minor pathway that involves the direct activation of the insulin receptor or insulin-like growth factor (IGF)-I receptor and a major pathway that acts through the inhibition of IGF binding proteins (in particular, IGFBP-1 and IGFBP-2), which may result in increased free and bioavailable IGF-I ( 41 ) . An important role of insulin and IGF-I in colorectal carcinogenesis is supported by in vitro studies, animal models, and epidemiologic studies ( 41 – 44 ) . Epidemiologic studies of circulating concentrations of insulin, C-peptide (a marker of insulin secretion), or IGF-I have shown two- to threefold increased risks of colorectal cancer for individuals in the highest exposure categories, compared with those in the lowest exposure categories ( 24 , 45 – 48 ) . Furthermore, a recent study reported that chronic insulin therapy was associated with a statistically significant increase in colorectal cancer risk among patients with type 2 diabetes ( 49 ) . Other mechanisms through which diabetes may be linked with the risk of colorectal cancer include slower bowel transit times in patients with diabetes, which could contribute to the increased exposure of colonic mucosa to potential carcinogens, and elevated concentrations of fecal bile acids associated with increased blood glucose and triglyceride concentrations ( 20 , 50 – 52 ) . Fecal bile acids have been shown to promote colorectal cancer in animal models ( 53 ) .

Our results have important clinical and public health implications. In the United States, about 8% of adults have diabetes ( 54 ) , and it has been predicted that the number of Americans with diagnosed diabetes will increase 165%, from 11 million in 2000 (prevalence of 4.0%) to 29 million in 2050 (prevalence of 7.2%) ( 55 ) . Colorectal cancer is the second leading cause of cancer death in the United States and other Western countries ( 2 , 3 ) . The prevalence of diabetes will probably increase as a result of the growing obesity epidemic, and thus this disease may contribute to the development of additional cases of colorectal cancers.

In summary, the results from this meta-analysis strongly support an association between diabetes and increased risk of colon and rectal cancer in both women and men. These findings provide evidence for a role of hyperinsulinemia or factors related to insulin resistance in colorectal carcinogenesis.

References