The main finding of our study was that diabetic patients with breast cancer receiving metformin and neoadjuvant chemotherapy have a higher pCR rate than do diabetic patients not receiving metformin (24% v
= .007; ). This was not due to a difference in the amount of chemotherapy delivered, because this was balanced between the two diabetic groups. The multivariate model shows that metformin use was independently predictive of pCR (odds ratio, 2.95; 95% CI, 1.07 to 8.17; P
= .04) after adjustment for diabetes status, BMI, age, stage, grade, ER/PR and HER-2 status, and neoadjuvant taxane use (). These results are consistent with epidemiologic data showing metformin use in diabetics decreases both cancer incidence and mortality.17,18
They are also consistent with the known inhibitory effect of metformin on the growth of cancer cell lines19–22
and of tumors in animal models.21–24
The mechanism of the antiproliferative effect of metformin is a matter of ongoing study. Type 2 diabetes is associated with obesity and metabolic syndrome. Patients with type 2 diabetes are insulin resistant and hyperinsulinemic.27
There is evidence to suggest that elevated insulin levels and the associated changes in levels of insulin-like growth factors, sex hormones, and adipokines contribute to tumorigenesis.3,13–16
Metformin partially reverses hyperinsulinemia and may also have antiproliferative effects via this mechanism. In clinical studies, elevated insulin levels have been associated with poorer outcomes in patients with breast cancer.28–30
Metformin has been shown to reduce insulin levels by 22% in nondiabetic hyperinsulinemic women with early-stage breast cancer.31
This effect may have been at play in our diabetic population, which was composed largely of overweight or obese patients with type 2 diabetes who were expected to have elevated insulin levels ().
With regard to this, there is recent evidence for the efficacy of nonpharmacologic interventions in reducing insulin resistance and possibly affecting breast cancer outcomes. For instance, women randomly assigned to 16 weeks of a strength and endurance exercise intervention showed decreases in fasting insulin levels and in insulin resistance.32
The effects of this intervention on outcome have not yet been reported. In the Women's Intervention Nutrition Study, in the hormone receptor–negative subset, lower dietary fat intake and weight loss in the intervention group correlated with lower long-term mortality compared with that of the control group (7.5% v
There is no insulin level analysis available, but the result is consistent with the notion that factors other than estrogen, such as insulin, are important in this ER- and PR-negative subset. However, some studies have reported no association between exercise interventions and changes in insulin levels in breast cancer survivors.35,36
Interestingly, in our study, the pCR rate was also lower in the nonmetformin diabetic group compared with that in the nondiabetic group (8% v
= .04). The rate of 16% in the nondiabetic patients is consistent with other studies of taxane-based neoadjuvant chemotherapy in patients with breast cancer.1,2,25
The lower rate in the nonmetformin group raises the possibility that other factors particular to this group make the tumors less susceptible to neoadjuvant chemotherapy. The higher rate in the metformin group compared with that in the nonmetformin group suggests these factors may be reversed by metformin.
We examined the possibility that exogenous insulin administration, which may promote tumorigenesis and which was different between the metformin and nonmetformin groups (16% v 33%), might have had an effect on the differences in pCR rates between these groups. There was no difference in the pCR rates in the metformin group (27% for insulin use v 23% for no insulin use; P = .75). However, insulin use was associated with a significant decrease in the pCR rate in the nonmetformin group (0% for insulin use v 12% for no insulin use; P = .05). These results suggest that part of the difference in pCR rates between the metformin and nonmetformin groups (and between the nonmetformin and nondiabetic groups) may be attributable to insulin use. However, insulin use is only one factor, because patients who were not receiving insulin also seemed to benefit from the addition of metformin (pCR rate increased from 12% to 23%).
Metformin activates the AMP-activated protein kinase (AMPK) pathway in a manner dependent on the upstream kinase LKB1. In hepatocytes, this results in inhibition of gluconeogenesis, and this is the principal mediator of the glucose- and insulin-lowering effects of metformin.37
Under low-energy and other stress conditions, AMPK phosphorylates a number of targets to inhibit cellular growth and proliferation, including components of the growth-promoting mammalian target of rapamycin pathway.38–40
Whether the apparent antitumor effect of metformin in our diabetic patients may have been mediated by endogenous insulin or insulin-like growth factors, AMPK, the mammalian target of rapamycin pathway, or other pathways remains to be defined.
An exploratory survival analysis conducted at a median follow-up of 37 months showed no significant difference in 3-year RFS between the three groups (P
= .66). However, there was a difference in 3-year OS; patients in the diabetic groups were doing worse than were those in the nondiabetic group (P
= .02). This is consistent with the known worse outcomes of diabetic versus nondiabetic patients with breast cancer.6–8
In addition, although the pCR rate in the metformin group was threefold that in the nonmetformin group, there was no significant difference in the RFS or OS between these two groups. This can be explained by the modest effect substantial differences in pCR may have on RFS and OS, because the number of patients with pCR is often not large enough to impact the survival of the group as a whole. For example, the National Surgical Adjuvant Breast and Bowel Project B-27 study of neoadjuvant chemotherapy in breast cancer showed no differences in RFS or OS, despite a doubling of the pCR rate in the taxane-containing group (26%) versus the nontaxane-containing group (13%).1
However, the patients in this study who did achieve pCR had a significant improvement in RFS and OS compared with those who did not.1
Similar data from a number of other studies have confirmed that pCR is correlated with improved RFS and OS; thus pCR has become a surrogate end point for survival in neoadjuvant studies of breast cancer.2,41
To our knowledge, our study has provided the first clinical evidence of the potential efficacy of metformin as an antitumor agent in breast cancer. It was based on one of the largest breast cancer neoadjuvant chemotherapy databases available, although the number of diabetic patients was modest because of the relatively low prevalence of diabetes. Several factors may have differed between the study groups and resulted in bias in the outcome measures. These include potential misclassification of diabetic patients (most were self-identified and taking diabetic medications) and diabetes control (the available A1c data suggested no difference). A number of patients were excluded because of incomplete medication or other records, and the reasons why certain patients were taking particular diabetic medications (such as metformin v insulin) are unknown. As with any retrospective study, there remains the possibility of unidentified confounders nonrandomly distributed between the groups of interest.
The main finding of the study—that there is an association between metformin use and higher pCR rates in diabetic patients receiving neoadjuvant chemotherapy—is hypothesis generating and consistent with the idea that metformin may have an antitumor effect in patients with breast cancer. In combination with the growing body of preclinical data, this suggests that this hypothesis deserves to be tested prospectively. We are conducting additional clinical and laboratory studies to evaluate the potential of this interesting and widely used diabetes drug as an antitumor agent.