|Home | About | Journals | Submit | Contact Us | Français|
Recent observational studies have shown that metformin use in diabetics decreases both cancer incidence and mortality. Metformin use is also independently predictive of pathologic complete response. We explored the association between metformin use and survival outcomes in patients with triple receptor-negative breast cancer (TNBC) receiving adjuvant chemotherapy.
The Breast Cancer Management System database of The University of Texas M.D. Anderson Cancer Center identified 1448 women who received adjuvant chemotherapy for TNBC between 1995 and 2007. Patients were categorized by diabetes status and metformin use. Kaplan-Meier product limit method was used to calculate distant metastasis-free survival (DMFS), recurrence-free survival (RFS), and overall survival (OS). Cox proportional hazards models were fit to determine the association between metformin use and survival outcomes.
Our study cohort consisted of 63 diabetic patients taking metformin, 67 diabetic patients not taking metformin, and 1318 non-diabetic patients. Patients in the diabetic groups tended to be older (P=0.005); more diabetic patients were postmenopausal (P=0.0007), black (P=0.0001), and obese (P < 0.0001). At a median follow-up of 62 months, there were no significant differences in 5-year DMFS (P=0.23), RFS (P=0.38), and OS (P=0.58) between the three groups. Compared to the metformin group, patients who did not take metformin (Hazard ratio [HR]=1.63; 95% CI:0.87 to 3.06; P=0.13) and nondiabetics (HR=1.62; 95% CI:0.97 to 2.71; P=0.06) tended to have a higher risk of distant metastases.
Our findings suggest that metformin use during adjuvant chemotherapy does not significantly impact survival outcomes in diabetic patients with TNBC.
It is believed that 8–18% of cancer patients are suffering from diabetes1. The increased coexistence of diabetes with breast cancer has been related to the mitogenic effects of signaling through the insulin/insulin-like growth factor-1 and insulin receptor2, 3. Diabetes mellitus is believed not only to be a moderate risk factor for breast cancer but, also, to considerably worsen prognosis in this disease4, 5.
Metformin is a widely prescribed oral hypoglycemic agent used as first-line therapy for type 2 diabetes mellitus. Several studies have suggested a possible association between the use of metformin and decreased risk of cancer6–9, and cancer-specific mortality in diabetic patients6, 10. The mechanisms involved in the anti-proliferative effects of metformin are probably very diverse11, 12. Preclinical studies with breast cancer cell lines demonstrate that metformin acts as a growth inhibitor which is mediated by upregulation of AMP-activated protein kinase (AMPK) activity and downstream suppression of signaling through the mammalian target of rapamycin (mTOR)13. Other mechanisms include, in particular, influences on estrogen biosynthesis and estrogenic signal transduction14–16, and suppression of human epidermal growth factor receptor-2 (HER2) protein expression17. More recently, Liu et al.18 demonstrated unique apoptotic effects of metformin against triple receptor-negative breast cancer (TNBC) cell lines via PARP cleavage and the activation of both intrinsic and extrinsic caspase signaling cascades. Moreover, in nude mice, metformin modestly inhibited tumor growth of xenografts of a TNBC cell line.
Recent observational studies have shown that metformin use in diabetics decreases both cancer incidence and mortality6, 10. Strikingly, this effect seemed to improve with higher doses6, 19. Despite these promising results, there is only sparse evidence from clinical studies addressing the relative impact of metformin on the breast cancer outcome. Furthermore, it is not clear whether metformin use is predictive of improved long-term survival outcomes.
The objective of this work was to investigate in a retrospective setting the distant metastasis-free survival (DMFS), recurrence-free survival (RFS) and overall survival (OS) outcomes among diabetic patients taking metformin, diabetic patients not taking metformin, and non-diabetic patients non diabetic patients.
The Breast Cancer Management System database of The University of Texas M.D. Anderson Cancer Center (MDACC) identified 1448 women who were diagnosed and treated with adjuvant chemotherapy for TNBC between 1995 and 2007. Patients with metastatic or bilateral disease, prior history of cancer, resolved gestational diabetes or diabetes diagnosed after the period of adjuvant chemotherapy were excluded from the analysis.
This study was approved by the Institutional Review Board at MDACC. Accuracy of clinical variables recorded within the prospectively collected data was verified by retrospective review of individual patient records.
All pathologic specimens were reviewed by dedicated breast pathologists at MDACC. Invasive carcinoma was confirmed on initial core biopsy specimens. Initial clinical stage and pathologic stage of all patients were revised and based on the seventh edition of the American Joint Committee on Cancer (AJCC) staging criteria20. Histologic type and grade were defined according to the WHO classification system21 and modified Black's nuclear grading system22, respectively. Negative estrogen-receptor (ER) and progesterone-receptor (PR) status was defined as nuclear staining of ≤ 10% on immunohistochemical (IHC) analysis. HER2-negative status was defined as either 1+ or no staining by IHC and/or absence of gene amplification by fluorescence in situ hybridization.
The type of surgery undertaken and adjuvant chemotherapy received were at the discretion of the patient and multidisciplinary treating team. Surgical intervention was breast conserving surgery (BCS) for 51% of patients (n=858) and mastectomy for 49% of patients (n=821). Adjuvant chemotherapy regimens comprised of anthracycline-based regimens with a taxane (n=758, 45%), or without a taxane (n=675, 40%); single-agent taxane (n=65, 4%); and non-anthracycline/non-taxane-containing regimens (n=181, 11%). Postoperative radiotherapy was administered if patients had BCS, locally advanced disease at presentation, primary tumor of greater than 5 cm, or equal or greater than 4 involved axillary nodes. None of the patients received adjuvant endocrine therapy.
The demographic and clinical characteristics of the three study groups were compared with the chi-square test. DMFS, RFS, and OS were defined as the time from the date of definitive surgery until the first date of documented distant metastasis, disease recurrence, or death from any cause, respectively. Patients not experiencing the relevant end points were censored at last follow-up. Survival outcomes were estimated with the Kaplan-Meier product-limit method; groups were compared with the log-rank statistic. Cox proportional hazards models were fitted to determine the association of the use of metformin with survival outcomes after adjustment for age (>50 vs. ≤50), tumor stage (T3–4 vs. T1–2), nodal stage (N1–3 vs. N0), nuclear grade (III vs. I/II), lymphovascular invasion (LVI) (positive vs. negative), and chemotherapy type (anthracycline-based vs. taxane-based vs. anthracyclines and taxane-based vs. other). Factors included in the multivariate models for adjustment were based on both statistical and clinical significance. Hazard rate and ratio values were estimated using kernel functions23. Statistical analyses were carried out using SAS 9.1 (SAS Institute Inc., Cary, NC) and S-Plus 7.0 (Insightful Corporation, Seattle, WA).
A total of 1448 patients were included in the statistical analysis, of whom 130 (9%) were diabetic and 1318 (91%) were non-diabetic. Of all the diabetic patients, 63 (48%) were on metformin during adjuvant chemotherapy (metformin group). In the non-metformin group, glycemic control was achieved by using different modes of anti-diabetic therapy, including dietary treatment only (30%), sulfonylurea preparations (28%), insulin therapy (39%), or thiazolidinediones (20%). Patient demographics, clinical characteristics including co-morbidities and concomitant medications are summarized in Table 1. Patients in the diabetic groups were older (P=0.005) than those in the non-diabetic group. Diabetic groups also had a higher proportion of postmenopausal (P=0.0007), black (P= 0.0001), and obese (P < 0.0001) patients than the non-diabetic group did. The other potential prognostic factors were not significantly different among the three groups. Aspirin was more frequently used in the non-metformin group than the metformin group (P=0.05); however there were no differences in the use of thiazolidinediones, statin, angiotensin converting enzyme inhibitors and angiotensin receptor blockers between the two diabetic groups.
The median follow-up for the patients that are still alive was 62 months (range 1–176 months). The Kaplan-Meier estimates of DMFS, RFS and OS stratified by groups (metformin, non-metformin and non-diabetic) are displayed in Figure 1A–C.
Of all patients, 559 (39%) experienced distant metastases; 18 (29%) in the metformin group vs. 26 (39%) in the non-metformin group vs. 515 (39%) in the non-diabetic group (Table 2). The 5-year DMFS estimates were 73%, 66% and 60% in the metformin group, the non-metformin group, and the non-diabetic group, respectively (P=0.23). When comparing patients who took metformin with patients who did not, the 5-year DMFS estimates were 60% vs. 73% (P=0.09). After adjustment for age, body weight, tumor size, nodal status, nuclear grade, LVI, and type of adjuvant chemotherapy, compared to the metformin group, patients who did not take metformin (Hazard ratio [HR]=1.63; 95% CI:0.87 to 3.06; P=0.13) and non-diabetics (HR=1.62; 95% CI:0.97 to 2.71; P=0.06) tended to have a higher risk of distant metastases (Table 3).
Overall, there were 647 (44%) recurrences; 24 (38%) in the metformin group, 29 (43%) in the non-metformin group, and 594 (45%) in the non-diabetic group. The estimated 5-year RFS rates were not different among the three groups (65% in the metformin group, 64% in the non-metformin group, and 54% in the non-diabetic group, P=0.38). After adjustment for age, body weight, tumor size, nodal status, nuclear grade, LVI, and type of adjuvant chemotherapy, there were no significant differences in risk of recurrence between the non-metformin (HR=1.37; 95% CI:0.78 to 2.4; P=0.27) vs. metformin groups; and the non-diabetic (HR=1.36; 95% CI:0.87 to 2.1; P=0.17) vs. metformin groups.
In univariate analyses, older age (>50), smaller tumor size, absence of metastatic lymph nodes or LVI, undergoing BCS rather than mastectomy, and receiving adjuvant radiotherapy or anthracycline-based chemotherapy were associated with higher DMFS and RFS rates. In multivariate analyses, older age, T1–2 status, N0 status, and negative LVI remained as independent significant predictors for improved DMFS and RFS.
There were a total of 535 (37%) deaths; 20 (32%) in the metformin group vs. 23 (34%) in the non-metformin group vs. 492 (37%) in the non-diabetic group. The 5-year OS estimates were 67% in the metformin group, 69% in the non-metformin group, and 66% in the non-diabetic group (P=0.58). After adjustment for age, body weight, tumor size, nodal status, nuclear grade, LVI, and type of adjuvant chemotherapy, using metformin group as reference, non-metformin group (HR=1.22; 95% CI:0.66 to 2.28; P=0.52) and non-diabetic group (HR=1.28; 95% CI:0.79 to 2.08; P=0.31) had a nonsignificant increased risk of death. When comparing diabetics on metformin vs. non-metformin, the P-values (log-rank) were 0.29 for DMFS, 0.053 for RFS and 0.80 for OS, respectively.
In univariate analyses, younger age (≤50), T3–4 status, N1–3 status, positive LVI, and higher nuclear grade were factors associated with increased risk of death. In the multivariate model, the above noted prognostic features maintained their unfavorable impact except for younger age. Within the diabetic groups, stratified analysis for body weight and daily metformin dose (0 mg vs. [500, 1000 mg] vs. >1000 mg) did not reveal a significant difference in survival outcomes.
Figures 1D–1I show Kernel estimates of the hazard functions and HR of distant metastasis, recurrence and death, respectively. The 3- and 5-year distant metastasis HR were 1.2 and 2.3 for non-metformin vs. metformin groups; and 1.5 and 1.4 for non-diabetic vs. metformin groups. The 3- and 5-year recurrence HR were 1.1 and 2.3 for non-metformin vs. metformin; and 1.4 and 1.3 for non-diabetic vs. metformin groups. The 3- and 5-year death HR were 0.9 and 1.3 for non-metformin vs. metformin groups; and 1.1 and 1.3 for non-diabetic vs. metformin groups.
Herein, we present the results of our single-institution study in a large cohort of women with TNBC, and assess the role of metformin during adjuvant chemotherapy in survival outcomes. Our data suggests that concurrent use of metformin with adjuvant chemotherapy does not significantly impact survival outcomes in diabetic patients with TNBC; however patients who do not take metformin and non-diabetics tend to have a higher risk of distant metastases.
In our study cohort, diabetic patients were more likely to be black and obese. While non-diabetic patients tended to be younger, they had similar clinical stage of disease at presentation compared to diabetics. Consistent with the previous findings24, 25, the tumor histopathological features were not different between the diabetic patients and non-diabetics. The choice of adjuvant chemotherapy did not differ between the three groups. We also evaluated the use of thiazolidinediones and other concomitant medications in the diabetic groups, as there is evidence that these - thiazolidinediones26, aspirin27, 28, statin29, 30, angiotensin converting enzyme inhibitors/ angiotensin receptor blockers31- may affect tumorigenesis. Among these medications, only aspirin use was significantly different between the two diabetic groups.
In contrast to previous studies25, 32–34, our findings showed that in this group of patients with TNBC, the presence of type 2 diabetes mellitus was not associated with lower survival outcomes even when diabetic patients had other poor prognostic features including black race and obesity35–39. We could speculate that worse survival in diabetic patients, is most likely ameliorated by strict glycemic control and therefore, through the restoration of tumor sensitivity to chemotherapy and radiotherapy.
At the level of cell signaling, metformin is known to improve insulin resistance-associated hyperinsulinemia mainly by affecting the insulin/insulin-like growth factor-1 (IGF1) signaling pathway which has been shown to induce cell cycle proliferation, tumor formation, and metastases40–42. Metformin also exerts direct inhibitory effects on PI3K/Akt/mTOR signaling pathway through activation of AMPK13, 43–45. Consequently, metformin was associated with a decreased risk of breast cancer in patients with type 2 diabetes mellitus in various observational studies6, 7. Besides, evidence suggests a lower cancer-related mortality rate in diabetic patients treated with metformin compared to sulfonylurea preparations and, possibly, insulin therapy6, 8, 10, 46, 47.
It is not clear whether metformin use is predictive of improved long-term survival in breast cancer. Epidemiological studies reported up to 23% reduced risk of cancer-related mortality for metformin users compared with that for nonusers10, 47. Jiralerspong et al. reported a three-fold greater pathologic complete response (pCR) rate in diabetic breast cancer patients who received metformin during neoadjuvant chemotherapy than those who did not (OR:2.95; 95% CI:1.07 to 8.17; P=0.04)25. With a limited follow up time, the estimated 3-year RFS and OS rates tended to be better in the metformin group vs. the non-metformin group. However, metformin was not an independent predictor of either RFS or OS after adjusting for diabetes status, BMI, age, stage, grade, ER/PR status, and taxane use. While in our cohort of TNBC, we did not see a significant survival benefit with concurrent use of metformin and adjuvant chemotherapy, in the multivariate survival model, patients on metformin tended to have a reduced risk of developing distant metastasis compared to patients who were not on the drug (P=0.06 when compared with non-diabetics). Additionally, when the two diabetic groups were compared (metformin vs. non-metformin), there was a beneficial effect of metformin use in diabetic patients for RFS (P=0.053).
Metformin has been shown to reduce insulin levels by 22% in non-diabetic hyperinsulinemic women with early-stage breast cancer48. In clinical studies, elevated insulin levels have been associated with an increased breast cancer recurrence and death49, 50. Further, there is recent evidence for the efficacy of non-pharmacologic interventions that reduce insulin resistance in affecting breast cancer outcomes. For instance, in the Women's Intervention Nutrition Study24, 51, lower dietary fat intake and weight loss in the intervention group correlated with lower relapse events (9.8% vs. 12.4%) and death (7.5% vs. 18.1%) compared with that of the control group. In the subgroup analyses, dietary fat reduction had a greater effect on relapse-free survival in women with ER-negative breast cancer tumors.
The known adverse prognostic factors, such as younger age52, 53, larger tumor size54, positive nodal status and LVI55–57 were confirmed to be prognostic in this analysis. We also demonstrated that patients treated with either single-agent taxane or non-anthracycline/non-taxane-containing chemotherapy regimens had an increased risk for distant metastases, recurrence, and death. The differential influence of anthracycline-containing chemotherapy regimens on breast cancer outcome of patients with TNBC tumors will require confirmation. Recently, Hirsch et al58 demonstrated that combinational therapy of metformin and doxorubicin reduces tumor mass and prevents relapse much more effectively than either drug alone in a breast cancer xenograft mouse model. Moreover, selective inhibition of breast cancer stem cells by metformin had a dramatic effect on prolonging remission58. This interesting evidence provides the rationale for studying metformin with an anthracycline-based chemotherapy regimen as a new treatment for breast cancer.
Although previous studies have shown potential efficacy of metformin as an antitumor agent in breast cancer, these studies differ from the present work in that no account of the adjuvant breast cancer treatment was presented. However, several limitations must be considered when interpreting the results of our study. It should be noted that our conclusions are mainly based on the subgroup of 63 metformin users and 1385 control patients, relying on a relatively small number of case and control patients. In addition, patient selection for individual treatment regimens may have influenced the differences in oncologic outcome. We did not have available data on duration of diabetes, metformin use before the index date in the diabetic group, or metformin use in the non-diabetic group for other comorbid conditions such as polycystic ovary syndrome and non-alcoholic fatty liver disease. Another limitation is that the level of surveillance and detection of nonfatal outcomes, in particular, might vary according to diabetes status. We also could not add the comorbidity score as a potential adjustment variable since we did not have enough information about the severity of comorbid conditions. Hence, the potential impact of these confounding factors on outcome need to be acknowledged.
In conclusion, the results of our retrospective analysis suggest that metformin use during adjuvant therapy was not associated with improved survival outcomes in diabetic patients with TNBC; however there was a trend towards a decrease in the risk of developing distant metastasis in diabetic patients taking metformin compared with non-diabetics. In light of the evidence that most of the breast cancer deaths result from distant metastatic relapses, these findings deserve to be tested with additional prospective studies. Currently, the impact of metformin on survival outcomes of patients treated with adjuvant breast cancer therapy is being investigated in a Phase III randomized trial conducted by national cancer Institute of Canada (NCIC) Clinical Trials Group.
Research support: This work was supported in part by National Cancer Institute 1K23CA121994-01, National Cancer Institute Breast Specialized Program for Research Excellence (Developmental Grant) P50-CA116199 (to AMG). The M.D. Anderson Breast Cancer Management System is supported in part by the Nelly B. Connally Breast Cancer Research Fund.
Disclosures: The authors have no financial interest to declare.