Serum glucose and risk of cancer: a meta-analysis – BMC Cancer

This is the first meta-analysis examining the association of serum glucose and cancer risk. We found a consistent positive association, which was not altered strongly by sex, study type, or cancer type.

As previously described, several molecular mechanisms have been postulated in an effort to explain the association between glucose and carcinogenesis. The insulin – IGF-1 axis is the most commonly suggested pathway [93]. Our results showed a weaker association for IGF-1 driven cancers than the overall association or non-hormonally driven cancers, suggesting that if the insulin- IGF- 1 axis does play a role it is likely to be as part of a more complex molecular mechanism.

Another proposed mechanism is an increased availability of sex hormones caused by a reduction of SHBG in the presence of hyperinsulinaemia [7, 94]. However, our meta-analysis showed a similar association between elevated serum glucose and risk of hormonally and non-hormonally driven cancers. This suggests that this is not the only underlying mechanism for the link between glucose and cancer. It is possible that other mechanisms, i.e. chronic inflammation [10–12] or direct actions of glucose [16], may also be playing a role.

To our knowledge this is the first comprehensive meta-analysis looking at epidemiological studies of serum glucose levels and cancer risk. Existing meta-analyses to date focused on the association between serum glucose levels and a specific type of cancer [4, 76]. A breast cancer-specific study including ten cohort studies found that the association between serum glucose levels and risk of breast cancer was small in non-diabetic subjects (pooled RR: 1.11 (95% CI: 0.98-1.25) [4]. The direction of this study is consistent with our findings, however our meta-analysis focused on high serum glucose levels as defined by the WHO definition for metabolic syndrome so that we also included potential diabetic subjects. Thus, when investigating serum levels of glucose, it is also important to consider diabetes. A bladder cancer-specific study showed that diabetes was associated with a 30% increased risk (95% CI: 1.18-1.43), which is consistent with the direction of the association found for serum glucose and cancer in our meta-analyses [76]. Other cancer types which also show a positive association with diabetes include pancreatic, endometrial, breast and colorectal cancer [20–22, 95], however an inverse association has been observed for prostate cancer [91]. The latter must be interpreted with caution as diabetics have higher morbidity and mortality from other diseases. There may be competing risks masking their risk of prostate cancer [96]. However, it is important to note that diabetes is a slightly different exposure than serum levels of glucose as diabetic treatments may normalise glucose levels and potentially also affect risk of cancer [43].

We made every effort to include all relevant publications available to date through various sources, including grey literature, and the two main online databases (Pubmed and Embase). We were able to also access unpublished data from the MECAN group enabling us to include this large cohort of over 500,000 subjects [18]. In addition, clearly defined objective criteria for exposure, outcome, and other study characteristics were specified a priori. One limitation of our study is the heterogeneity between the different categorization methods for glucose ranges across the include studies. We tried to overcome this by combining the different categories as similarly as possible and believe this cannot significantly affect our findings. Nevertheless, this made it not possible in the current meta-analysis to make a distinction between pre-diabetes and diabetes. The overall results showed a rather large amount of heterogeneity, as suggested by the I2 statistic. All of our sensitivity and subgroup analyses showed consistent findings in terms of direction of the association, while the heterogeneity remained high. Only when we conducted tumour specific analysis, the I2 statistic reduced. This suggests that heterogeneity is most likely explained by combining studies with different outcomes. However, the consistent finding of a positive association in all our analyses supports the robustness of our findings. Six of the studies included, either had mixed or did not specify fasting status. However, exclusion of these studies did not alter the association observed. A further limitation is the lack of information regarding the diagnosis of diabetes, use of oral hypoglycaemics or insulin in those included in the studies. Future research including adjustment for components such as age, cancer treatment, diabetes (or its treatments), or BMI would be useful in confirming the importance of raised glucose in carcinogenesis. All studies included were soundly designed and executed epidemiological studies, which clearly defined their methodology. However, the size of the studies did vary considerably. The two largest studies [5, 18] did account for well over half of the cases included, but they represent a Korean and European population which we believe can be broadly applicable to all patient populations. Limitations reported by the individual studies overlap widely. They include, having only localised cancer as an outcome, small sample size, specific demographic groups only (i.e. smokers only), lack of information on diabetes and obesity and all but one study [81] used single measurements of glucose for their analysis.