Hyperglycemia, Obesity, and Cancer Risks on the Horizon

The prevalence of diabetes mellitus has increased substantially over the past several decades, in part because of the growing epidemic of obesity.¹ Approximately 8% of the US population has diabetes,² with more than 90% of cases classified as type 2 diabetes, and with insulin resistance as the major underlying pathophysiology. Much of the emphasis in diabetes management focuses on reducing end-organ complications, including retinopathy, neuropathy, nephropathy, and macrovascular disease. Several recent epidemiological studies, including the report by Jee et al³ in this issue of JAMA, have also shown an association between diabetes and several common cancers, including pancreas and colon cancer. How important is this effect and what are the public health implications?

The prospective cohort study of nearly 1.3 million people conducted by Jee et al³ took advantage of data collected by the insurance provider for all government employees, teachers, and their dependents in Korea. A cohort of individuals was enrolled over 4 years at the time of required biennial medical examinations and was then followed up for as long as a decade. Outcomes were measured through national registration of death certificates, a national cancer registry, and hospitalization records. The main finding was that elevated fasting serum glucose levels were associated with a 27% increase in cancer mortality among men and a 31% increase among women. A dose-response trend was evident, with higher fasting glucose levels associated with higher cancer mortality rates. Cancer incidence rates followed similar patterns.

One of the unique aspects of this study³ is that Korea has a low rate of obesity in the population. The average body mass index (BMI) in the study population was 23.2, with approximately one fourth of the population considered overweight (defined as BMI ≥25.0). By comparison, recent National Health and Nutrition Examination Survey data demonstrate that approximately 65% of US men and women are overweight or obese, nearly one third are obese (defined as BMI ≥30), and 5% are extremely obese (defined as BMI ≥40).⁴ If current trends continue, 40% of the US population will be classified as obese in 2010.⁵ In the United States, diabetes is tightly linked to obesity, and the proportion of individuals with diabetes who meet the definitions of overweight and obese is higher than in the general population (79% and 46%, respectively).⁶ For the first time, this Korean study has made significant progress toward separating the specific effects of elevated blood glucose from the compendium of other metabolic abnormalities observed in obesity. The investigators were able to demonstrate a direct relationship between fasting serum glucose levels and cancer mortality across all BMI groups (<20, 20 to <23, and >23).

A wealth of epidemiological data supports the association between obesity and various types of malignancies, including postmenopausal breast cancer and cancers of the colon, pancreas, endometrium, esophagus (adenocarcinoma), kidney, gallbladder, and gastric cardia.⁷ The link between obesity and cancer may be mediated by insulin resistance and hyperinsulinemia. However, several complex physiological changes also result from obesity, including alterations in sex steroid levels. For example, obesity is associated with increased levels of bioavailable estrogen due to enhanced biosynthesis in adipocytes and reduced sex hormone–binding globulin levels.⁷ Thus, in the complex setting of obesity, it is difficult to tease out the specific molecular alterations that lead to cancer, and these contributions are likely to be site-specific.

Well-designed cohort studies can provide compelling observational data, and the study by Jee et al³ has several methodological strengths. First, it is likely that the cohort is reasonably representative of Korea, since the insurance provider covers more than 10% of the Korean population, and only 2.1% of the participants were excluded on the basis of incomplete data. Second, important information about other risk factors was collected by these investigators through use of a lifestyle and medical history questionnaire. In contrast, many other population-based cohort studies rely on national linkage registries that do not have this type of information, thereby limiting the ability to adjust for potential confounders. Third, the study cohort is extremely large, which increases the ability to recognize and distinguish very small relative risks as statistically significant.

This last attribute is important, because small risks can be meaningful. Tangible elevations in cancer risk were observed in Korean men and women who had fasting serum glucose measurements that were below the commonly recognized threshold of 125 mg/dL for a diagnosis of diabetes. Indeed, a fasting glucose level from 90 to 109 mg/dL was associated with a 4% increase in risk of cancer mortality among men (hazard ratio, 1.04; 95% confidence interval, 1.01-1.07). Even though this relative risk is minuscule, what it reflects in the broader context of this study is that elevated fasting glucose levels confer risk on a continuous scale. Just as clinical thresholds for hypertension and serum lipids are not easily categorized into “healthy” and “diseased,” the same appears to be true for fasting serum glucose.

Of course, no observational study is perfect, and this study has its limitations. The strongest relative risks were associated with pancreatic cancer, and despite the appropriate methodological approaches used in this study, the possibility that hyperglycemia was caused by the pancreatic cancer through endocrine insufficiency cannot be excluded definitively. In addition, since this study is not randomized, the exposed and unexposed groups might not have been appropriately balanced in the analysis. It is also possible that individuals with hyperglycemia were followed up more carefully by physicians, which would lead to more cancer diagnoses. However, this would not explain the strong findings with respect to cancer mortality. Moreover, cancer is not a single disease, and the case for hyperglycemia as a risk factor for “all cancer” is driven primarily by the data from the digestive cancers among men.

Among women, the summary result is related primarily to risks measured for pancreatic cancer. While this study did not describe the risks of endometrial cancer associated with hyperglycemia, both diabetes and obesity are recognized risk factors for this disease.⁸ Similarly, the analysis of breast cancer leaves more questions than answers. The modest elevation in breast cancer incidence with elevated fasting blood glucose observed by Jee et al³ was not statistically significant, although addition of women identified through use of medication for diabetes was associated with a significantly increased risk (hazard ratio, 1.51; 95% confidence interval, 1.26-1.80). These findings raise questions about how hyperglycemia might operate differently in premenopausal and postmenopausal women and suggest that various aspects of diabetes treatment should be addressed in future analyses of the data.

How may these results be interpreted in the context of cancer incidence and mortality in the population? Fortunately, the relative risks are modest and, therefore, the fraction of cancers attributable to elevated fasting glucose in the Korean population is small, in part because of the relatively low prevalence of diabetes in Korea (5%). Based on data provided in this study, an estimated 3.9% of cancer deaths in men and 0.8% of cancer deaths in women are attributable to diabetes, representing 846 of the total 26 473 cancer deaths observed in this study. However, these numbers may not reflect unmeasured confounding or effect modification.⁹ Some of these cancer deaths may be preventable, and further studies will be required to demonstrate that reductions in hyperglycemia and diabetes will lead to declines in cancer mortality. This is especially true in light of data from the Diabetes Prevention Program Research Group demonstrating a reduction in the incidence of diabetes through lifestyle interventions, including dietary modification and introduction of a regular exercise routine.¹⁰ Since the hazard ratios reported by Jee et al³ are derived from a Korean population, they cannot be directly implemented to increase understanding of the population-attributable fraction of cancer that may be due to diabetes in the United States. But since the prevalence of diabetes is higher in the United States than in Korea, it is possible that preventing diabetes may have a more important effect in the United States.

As diabetes becomes an increasing public health concern in modern societies, the cancer risks looming on the horizon are now being recognized. Strategies to address the emerging epidemics of diabetes and obesity are likely to have a broad impact on public health. Indeed, these approaches may ultimately diminish the burden of cancer for future generations.