Diabetes and cancer: Epidemiology and potential mechanisms

Abstract

Type 2 diabetes mellitus is an independent risk factor for cancer such as pancreatic, liver, colorectal and breast cancer. In addition, diabetes decreases the risk of prostate cancer. These associations have been found in numerous epidemiological studies, among them several prospective cohorts. However, such studies do not prove causality of the association and cannot exclude inadequate correction for known confounders (e.g. visceral fat) or the influence of unknown confounders. Thus, it is unclear whether the cancer risk is increased by the causes (e.g. the metabolic syndrome), the metabolic consequences (e.g. hyperglycaemia) or the therapy of diabetes; a question which is of major importance for therapeutic guidelines. Identification of the potential mechanisms by which the diabetes or its therapy accelerates or inhibits the development of cancer will help answering the question by providing biological plausibility. This review will summarize the evidence supporting the association of cancer and type 2 diabetes and discuss its potential causes.

Introduction

The association between diabetes mellitus and cancer has been recognized by clinicians for more than 80 years. In a survey of approximately 10,000 patients with diabetes mellitus treated at the Joslin clinic, 256 cases of cancer were identified. ¹ Of these cancers, 13% were of the pancreas, a proportion that is much higher than observed in non-diabetics. This association between diabetes and pancreatic cancer was later confirmed in numerous publications,^2–4 among them a large prospective Swedish cohort study which determined alterations of the cancer risk 10 or more years after the diagnosis of diabetes, thereby excluding reverse causality. ⁵ In addition, a large body of evidence has been accumulated from epidemiological studies indicating that other types of cancer are also diagnosed more frequently in diabetic patients (recently reviewed by Shikata ⁶ ). Risk increases in diabetics are highest for liver, pancreas and endometrial cancer (about twofold) and are lower but consistently observed for colorectal, breast and bladder cancer (1.2–1.5 fold). There is no association between diabetes and lung cancer and an inverse association between diabetes and risk of prostate cancer. Thus, elevated risk of some types of cancer can be considered a secondary complication of diabetes mellitus and/or of its metabolic and hormonal imbalances.

While the epidemiological evidence for the association of the two diseases is solid, there are a number of relevant, unsolved issues which were discussed in a recent consensus statement. ⁷ Is the development of cancer due to the metabolic abnormalities of diabetes (in particular the hyperglycaemia)? Or, is the elevated cancer risk a consequence of the metabolic syndrome (visceral obesity, dyslipidaemia, insulin resistance, hypertension) which causes diabetes and, independently, the development of cancer? The latter conclusion is supported by data from a prospective study indicating that normoglycaemic individuals who are at increased risk of diabetes already have an elevated cancer risk [hazard ratio (HR) = 1.7 ⁸ ]. Furthermore, the possibility has to be considered that the increase in cancer risk is an undesired side effect of the diabetes therapy.

It is evident that the answers to these questions are of major importance for prevention of diabetes-associated cancer. In this review, I therefore will summarize the available evidence for the association and causal relationship between diabetes mellitus and cancer and discuss alternative causes including obesity, the metabolic syndrome and a potential role of the diabetes therapy.

Methodological limitations

Epidemiological studies have important limitations inherent to the study design. Retrospective case–control studies are biased by the different selection of participants which usually causes an enrichment of individuals with healthier lifestyle in the control groups. By the nested case–control design of prospective cohorts, the selection bias is excluded; therefore, this study design is the preferred approach to prove associations. Unfortunately, most meta-analyses studying the association of diabetes with cancer include both retrospective case–control studies and prospective cohorts. Furthermore, a particular limitation of all observational studies is the possibility of confounding, where the confounder affects both an associated variable and an endpoint independently. If the confounder is known, the data can be corrected accordingly. However, residual confounding is still possible, as well as the presence of an unknown confounder. Thus, neither retrospective nor prospective case–control studies can prove causality of an exposure. Ultimate proof of causality therefore requires intervention studies which should at least monitor intermediate endpoints. In addition, the identification of a pathophysiological or pathobiochemical mechanism could provide biological plausibility for a causal relationship.

Essentially all studies investigating the association of diabetes with cancer do not distinguish between individuals with type 1 and type 2 diabetes. Since the prevalence of type 2 diabetes is approximately 20-fold higher than that of type 1 diabetes, it appears safe to conclude that the results reflect associations with type 2 diabetes and that no conclusion can be drawn with regard to type 1 diabetes.

Pancreatic cancer

A recent meta-analysis of 35 cohort studies indicated that diabetes mellitus is associated with an increased risk of pancreatic carcinoma [summary of relative risk (RR) = 1.94], with significant evidence of heterogeneity among these studies. ⁹ Contribution of reverse causality (with the carcinoma causing diabetes) to the association could be essentially excluded since an elevated risk (RR = 1.79) was also found in patients who were diabetic for at least 5 years. ⁵ Subgroup analyses revealed that the increased risk was independent of the body mass index (BMI), suggesting that the contribution of diabetes is independent of the metabolic syndrome. Furthermore, elevated levels predicted pancreatic cancer independently of obesity and insulin resistance, ¹⁰ also suggesting that the increased cancer risk is a consequence of the diabetes rather than of the metabolic syndrome. However, a meta-analysis of prospective studies also indicated an independent association between BMI and risk of pancreatic cancer in men and women. ¹¹ Thus, at least a part of the association between diabetes and pancreatic cancer could be due to confounding by adiposity. Potential mechanisms of these associations are a tumourigenic effect of hyperglycaemia, a mitogenic effect of the obesity-associated hyperinsulinaemia and a chronic, subclinical inflammation caused by fat infiltration of the pancreas. However, there is little direct evidence for these mechanisms. To the contrary, an intensified glucose lowering therapy with higher insulin doses has not led to an increased cancer incidence in several major clinical trials. ¹²

Liver cancer

The association between diabetes and liver cancer was first observed in a Scottish retrospective case–control study. ¹³ Subsequently, data from a Swedish prospective cohort convincingly demonstrated that diabetes mellitus is a major risk factor for hepatocellular carcinoma. ¹⁴ Recently, a meta-analysis of 18 cohort studies determined an about twofold higher risk of hepatocellular carcinoma in diabetic patients, independent of alcohol consumption or viral hepatitis status. ¹⁵ In a prospective study of 578,000 individuals, it has been shown that the risk is associated with each of the components of the metabolic syndrome (BMI, blood pressure, blood glucose, serum triglycerides), suggesting that not the diabetes per se (hyperglycaemia) but rather its metabolic causes are responsible for the association. ¹⁶ Indeed, obesity has been shown to be an independent and strong predictor of hepatocellular carcinoma. ¹⁷ As a potential mechanism, it has been suggested that fatty liver (non-alcoholic fatty liver disease (NAFLD)) leads to hepatosteatitis (non-alcoholic steatohepatitis (NASH)) which can progress to liver cirrhosis, the latter being a causal factor in the development of hepatocellular carcinoma. ¹⁸ The finding that fatty liver is an independent risk factor for the hepatocellular carcinoma is consistent with this assumption. However, in a comparison of 2061 patients with hepatocellular carcinoma with 6183 controls, the effect of diabetes on the risk of cancer remained significant after exclusion of patients with fatty liver disease. ¹⁹ Thus, it cannot fully be excluded that the metabolic abnormalities of diabetes (hyperglycaemia) or the metabolic syndrome (hyperinsulinaemia) independently contribute to the risk of liver cancer.

Colorectal cancer

Numerous studies have shown an association between type 2 diabetes and colorectal cancer. A meta-analysis of 15 studies (6 case–control and 9 cohort studies) including more than 2.5 million patients indicated an increased risk (summary RR = 1.3) of diabetic individuals compared with non-diabetic individuals. ²⁰ This association may reflect confounding by variables of exposure that independently predispose to both colon cancer and type 2 diabetes, such as visceral fat, physical activity, smoking, a high consumption of red meat or a low intake of dietary fibre. When the meta-analysis was restricted to studies that controlled for physical activity and BMI, a significant positive association between diabetes and colorectal cancer (summary RR = 1.34) was found. Furthermore, in a study of 48,000 Swedish men, the positive association between diabetes and colorectal cancer (RR = 1.45) was unaltered after adjusting for potential confounders (BMI, recreational physical activity, smoking status, education, family history of colorectal cancer, multivitamin supplement use, aspirin use and consumption of fruits, vegetables, dairy foods and red meat). ²¹ These data suggest that the specific abnormalities of diabetes mellitus, for example, hyperglycaemia, exert an independent effect on the risk of colorectal cancer and might play a causal role. However, it should be noted that the increase in risk by diabetes (RR = 1.3) is almost identical to that exerted by an increase in visceral fat (RR = 1.4). ²² Thus, it is striking that a correction for obesity (BMI or waist circumference) did not modify the association between diabetes and cancer risk. Consequently, residual confounding should not be excluded and might be responsible for the remaining risk.

In the absence of direct evidence linking the metabolic abnormalities of diabetes with the development of cancer, a causal role of diabetes remains unclear. As a mechanistic link between type 2 diabetes and the development of colon cancer, serum insulin and insulin-like growth factor I (IGF-I) levels have been proposed.^23,24 In type 2 diabetes, insulin resistance may lead to very high serum insulin levels and to lower levels of IGF binding proteins, resulting in an increased mitogenic activity of the serum. In addition, it has been suggested that the slower bowel transit time in diabetic patients could enhance the exposure of the colorectal epithelium to carcinogens (e.g. bile acids, nitrosamines, polycyclic hydrocarbons). ²⁵

Breast cancer

Numerous studies have demonstrated an association between type 2 diabetes and breast cancer; a meta-analysis of 20 studies (5 case–control and 15 cohort studies) indicated a risk increase of approximately 20%. ²⁶ The association was similar in studies that controlled for BMI, physical activity and alcohol consumption and in studies that did not adjust for these variables. Nevertheless, it cannot be excluded that a substantial portion of the association is due to a confounding effect of total body fat (BMI). The association between BMI and postmenopausal breast cancer has been analysed in the prospective European Prospective Investigation into Cancer and Nutrition (EPIC) cohort. ²⁷ Obese women (BMI > 30) had a 31% excess risk compared to women with BMI < 25. Hormone replacement therapy with estrogen derivatives increased the breast cancer risk in lean postmenopausal women, and no further increase by obesity was observed. ²⁷ Thus, it appears reasonable to assume that the association of BMI with breast cancer risk reflects a causal role of adipose tissue as a producer of estrogen. Consequently, body fat is a confounder of the association between diabetes and breast cancer, and it is so far unclear whether there is a residual, causal effect of the metabolic abnormalities of diabetes.

Prostate cancer

Prostate cancer is unusual in that it is inversely associated with diabetes mellitus: diabetic patients have a lower risk of prostate cancer than non-diabetic controls. This association is supported by a meta-analysis of 19 studies (7 case-control and 12 cohort studies) resulting in a summary RR of 0.84. ²⁸ The risk reduction was even higher in studies that adjusted for potential confounders such as BMI (summary RR = 0.74). It should be noted, however, that the analysed studies were very heterogeneous, with only 7 reporting a significant reduction, 1 reporting a significant increase and 11 observing no alteration of risk. A potential mechanism of this protective effect of diabetes is the lower levels of testosterone associated with diabetes mellitus. ²⁴

Diabetes therapy and cancer risk

Metformin

Observational studies^29–31 as well as the randomized clinical trial UK Prospective Diabetes Study (UKPDS) ¹² have suggested that patients treated with metformin had a lower risk of cancer than those on insulin or glyburide drugs. However, a meta-analysis of 14 studies failed to detect an effect of metformin on the overall cancer risk, ³² whereas a second meta-analysis of 15 studies calculated a significant effect of metformin on the colon cancer risk. ³³ Thus, at present no definitive conclusion as to a beneficial effect of metformin on the cancer risk can be made. However, it appears reasonable to conclude that metformin does not increase risk and is therefore a safe drug with regard to cancer risk.

Insulin therapy

As early as 1970, it was suggested ‘that the carcinogenic potentialities of exogenous animal insulins be investigated, particularly in view of their known antigenic and teratogenic properties in laboratory animals’. ³ Indeed, the insulin analogue AspB10 has been shown to induce spontaneous mammary tumours in rats. ³⁴ This effect appeared to be due to an extraordinarily high affinity to the IGF-I receptor, and studies to explore the therapeutic potential of the analogue were stopped. More recently, the human insulin analogue glargine was suspected to cause cancer, based on a retrospective analysis of data from health insurers. ³⁵ However, the effect depended on a widely criticized adjustment of the data ³⁶ and was not reproduced in other retrospective studies^37,38 or in randomized clinical trials.^39–41 Furthermore, higher insulin doses have not led to an increased cancer incidence in all major clinical trials investigating the effects of an intensified glucose lowering therapy. ¹² Thus, there is no convincing evidence for a carcinogenic role of any insulin derivative currently used in therapy. However, it cannot fully be excluded that the insulin therapy contributes to a hypothetical mitogenic effect of endogenous insulin. If proven, such a scenario would require a general re-evaluation of therapeutic strategies, and future research should address this important question.

Incretin-based therapies

Based on an analysis of the US Food and Drug Administration’s database of reported adverse events, the safety of the glucagon-like peptide-1 (Glp-1)-analogue exenatide as well as that of the dipeptidyl peptidase-4 (DPP4)-inhibitor sitagliptin has been questioned. ⁴² In this analysis, both agents gave rise to an approximately twofold increase in the risk of pancreatic carcinoma. However, since such data on adverse events are heavily biased (surveillance bias), the calculation of RRs is unreliable. So far, there is no confirmation of such an adverse effect from a prospective cohort or an intervention study. A retrospective analysis of health-care insurance data ⁴³ as well as a meta-analysis of 25 studies ⁴⁴ failed to confirm the association of exenatide with pancreatic cancer. Thus, the available data from human studies are neither sufficient nor robust enough to prove an adverse effect of the incretin-based therapy on cancer development.

Concluding remarks

A large body of evidence from observational studies indicates that type 2 diabetes is associated with an increased incidence of cancer of the pancreas, liver, endometrium, colon, breast and bladder. Although the epidemiological methodology suggests an independent effect of the diabetes (hyperglycaemia), confounding by obesity and hyperinsulinaemia cannot fully be excluded. Thus, in order to prevent diabetes-associated cancer, other potential mechanisms should also be considered: the total body fat and its estrogen production (breast cancer in women and men), total body fat and fat-induced pancreatic inflammation (pancreatic cancer), visceral fat and hepatosteatosis (hepatic cancer), and the metabolic syndrome and insulin resistance producing a mitogenic effect of insulin (all cancers). Efforts to prevent diabetes-associated cancer should target this pathophysiology and not exclusively focus on the control of hyperglycaemia. In addition, at present, it cannot be excluded entirely that the diabetes therapy plays a role in modifying the cancer risk. There is some epidemiological evidence suggesting that metformin therapy reduces the risk of diabetes-associated cancer. In contrast, a carcinogenic effect of insulin or insulin secretagogues remains hypothetical so far. Similarly, the published data on the incretin-based therapies are insufficient to support the concern of a carcinogenic effect and do not warrant an alteration of the therapeutic practice. Future studies of the diabetes therapy should consider cancer a secondary complication of type 2 diabetes and should aim at monitoring this clinical endpoint.