This article was published online on 20 October 2011. An error was subsequently identified. This notice is included in the online and print versions to indicate that both have been corrected 2 March 2012.
Metformin use has been associated with decreased cancer risk and mortality. However, the effects of metformin on clinical outcomes of colorectal cancer (CRC) are not defined. This study aimed to evaluate the association between metformin use and mortality of CRC in diabetic patients. We identified 595 patients who were diagnosed both CRC and diabetes mellitus. Patients were compared by two groups; 258 diabetic patients taking metformin and 337 diabetic patients not taking metformin. Patient's demographics, clinical characteristics, overall mortality and CRC-specific mortality were analyzed. After a median follow-up of 41 months, there were 71 total deaths (27.5%) and 55 CRC-specific deaths (21.3%) among 258 patients who used metformin, compared with 136 total deaths (40.4%) and 104 CRC-specific deaths (30.9%) among 337 patients who did not use metformin. Metformin use was associated with decreased overall mortality (p = 0.018) and CRC-specific mortality (p = 0.042) by univariate analysis. After adjustment for clinically relevant factors, metformin use showed lower risk of overall mortality (HR, 0.66; 95% CI 0.476–0.923; p = 0.015) and CRC-specific mortality (HR, 0.66; 95% CI 0.45–0.975; p = 0.037) in CRC patients with diabetes. Metformin use in CRC patients with diabetes is associated with lower risk of CRC-specific and overall mortality.
Colorectal cancer (CRC) is the third leading cause of cancer-related death.1 Despite major advances in surgical techniques and adjuvant chemotherapy, only a modest improvement in the survival of patients with advanced CRC has been achieved. Thus, new strategies for preventing and improving survival among CRC patients are required. Both CRC and type 2 diabetes mellitus (DM) generally occur in the elderly and share many risk factors such as physical inactivity, obesity and high saturated fat diet. Therefore, many patients are co-diagnosed with these conditions.
Although there has been some debate regarding the effects of diabetes on CRC, meta-analyses consistently show that type 2 DM is an independent risk factor for CRC,2, 3 and that diabetic patients with CRC may have worse outcomes than their non-diabetic counterparts.4 The pathophysiological mechanisms of diabetes, including insulin resistance, hyperglycemia and resulting hyperinsulinemia, are all associated with the development and progression of cancer.5 Moreover, some glucose-lowering therapies themselves are also associated with cancer risk and mortality, including CRC.6–9
Metformin, a biguanide derivative, is an oral drug widely used as a first-line therapy for type 2 DM. Metformin inhibits hepatic glucose production and decreases insulin resistance in peripheral tissues, thereby reducing levels of circulating glucose levels and improving insulin sensitivity. The molecular mechanism of metformin involves liver kinase B1 (LKB1)-dependent activation of adenosine monophosphate-activated protein kinase (AMPK).10 Activated AMPK inhibits the mammalian target of rapamycin (mTOR)-controlled synthesis of key proteins responsible for the malignant phenotypes of cancer cells as well as angiogenesis, and is regarded as potential anti-cancer mechanisms of metformin.11 Interestingly, several recent preclinical studies in rodent models have shown that metformin induces AMPK activation and inhibits tumor development and growth, including colon carcinogenesis.12–15 In addition, population studies have shown that patients with type 2 DM who are taking metformin have a lower risk of cancer and better outcomes compared with patients who do not take metformin.6, 7, 16, 17 Although there has been substantial evidence from in vivo and in vitro studies supporting the potential efficacy of metformin as an anti-cancer agent, there have been no clinical studies investigating the effect of metformin on CRC.
Therefore, we hypothesized that the anti-tumor activity of metformin would decrease the mortality rate of CRC. We undertook a retrospective cohort study to evaluate the effect of metformin on the clinical outcomes of CRC in diabetic patients.
Patients and Methods
We gathered 6,108 consecutive patients who were newly diagnosed with CRC between January 2000 and December 2008 at Severance hospital, Yonsei University, Seoul, Korea. Among them, 676 had type 2 diabetes, 81 of which were excluded based on the following exclusion criteria: diabetes diagnosed after CRC diagnosis (45), incomplete records (including medication records) (28), metformin use for less than 6 months (3) and any cancer previous to CRC diagnosis (5). Finally, 595 patients were analyzed and divided into two groups: 258 diabetic patients taking metformin and 337 diabetic patients not taking metformin. The use of other diabetic medications (sulfonylureas, thiazolidinediones, α-glucosidase inhibitors, insulin, etc.) and aspirin was also investigated.
Patient demographics and clinical characteristics, including age at diagnosis, gender, follow-up duration, duration of diabetes, family history of colorectal malignancy, body mass index (BMI), smoking history, drinking history and comorbidity (hepatitis, pulmonary tuberculosis and cardiovascular disease, including hypertension, coronary heart disease and cerebrovascular disease), were obtained from medical records. Laboratory findings included pretreatment carcinoembryonic antigen (CEA) levels, plasma glucose levels and hemoglobin A1c levels, as well as information pertaining to the CRC diagnosis, including stage, site, histology, differentiation, resection margin, lymphovascular invasion, microsatellite instability (MSI) status and treatment modality. The date of diagnosis of CRC was the day of pathologic diagnosis. All enrolled patients had undergone colonoscopic examination. We identified deaths through medical records, and determined the cause of death in all cases.
The institutional review board of Severance Hospital, Yonsei University, Seoul, Korea approved this study.
Tumor staging and treatment assessment
All patients were assessed by the 6th version of the AJCC staging system derived from a synthesis of clinical, pathological and imaging information. Treatment modality was decided by extension and location of tumor. Early localized or T2 N0 tumors were removed by endoscopic resection or surgery only. Locally advanced tumors or advanced tumors with resectable metastatic lesions were treated by surgery followed by adjuvant chemotherapy with or without radiotherapy, or by neoadjuvant chemotherapy or chemoradiation therapy following surgery. Advanced CRC with distant metastasis was treated by palliative chemotherapy or conservative care.
The primary endpoints of the current study were overall survival and CRC-specific survival. The baseline demographics and characteristics of patients were analyzed with descriptive statistics. Continuous data were analyzed using Student's t-test, and categorical data were compared by Pearson's χ2 test. Kaplan–Meier analysis was used for univariate analysis of overall mortality and CRC-specific mortality, and the survival curves of each group were compared by a log-rank test. A multivariate Cox proportional hazard regression model was used to evaluate overall mortality and CRC-specific mortality adjusting for various confounders, including age at diagnosis, sex, stage of cancer, BMI, diabetes duration, smoking status, HbA1c level, use of aspirin and use of insulin, sulfonylurea and thiazolidinediones. A value of p < 0.05 was considered significant. All statistical analyses were performed using SPSS version 13.0 (SPSS, Inc., Chicago, IL).
Patient demographics and clinical characteristics
Patient demographics and clinical characteristics are summarized in Table 1. The median age of patients was 63 years (range, 30–88 years). Baseline characteristics including age at diagnosis, sex, BMI, smoking history, drinking history, comorbidity, aspirin use, tumor stage, tumor site (colon or rectum), tumor histology, tumor differentiation, resection margin positivity, lymphovascular invasion, MSI status and pretreatment CEA level were not significantly different between the metformin group and the non-metformin group. Severity of diabetes as judged by hemoglobin A1c levels, pre-meal glucose levels and duration of diabetes were not different between the two groups. We also evaluated the use of other diabetic medications, because there is evidence that these medications, including insulin, sulfonylureas and thiazolidinediones, may affect tumorigenesis and prognosis.6, 8, 18 Insulin use was lower in the metformin group than in the non-metformin group (8.1% vs. 16.6%; p = 0.003), whereas thiazolidinedione use was higher in the metformin group than in the non-metformin group (8.9% vs. 3.9%; p = 0.014). Sulfonylurea use was not significantly different between the two groups. Meanwhile, a comparison of the treatment modality against CRC in the two groups showed no significant difference.
Table 1. Patient demographics and baseline clinical characteristics
Metformin use and survival analysis
The median follow-up duration was 41 months (range, 1–119). In the entire cohort, there were a total of 157 (26.4%) recurrences, 207 (34.8%) total deaths, and 159 (26.7%) CRC-specific deaths. With respect to metformin use, there were 71 total deaths (27.5%) and 55 CRC-specific deaths (21.3%) among 258 patients who used metformin, compared with 136 total deaths (40.4%) and 104 CRC-specific deaths (30.9%) among 337 patients who did not use metformin. The estimated 3-year CRC-specific survival rates were 92.4% and 90.8% for the metformin and non-metformin groups, respectively, and were significantly different (log rank p = 0.042) (Fig. 1a). The estimated 3 year overall survival rate was 89.6% for those patients who used metformin, compared with 87.9% for those who did not (log rank p = 0.018) (Fig. 1b).
Tumor stage (HR, 9.01; 95% CI 5.814–13.889; p < 0.001) and metformin use (HR, 0.66; 95% CI 0.45–0.975; p = 0.037) were independent predictive factors for CRC-specific mortality in diabetic patients (Table 2). After adjustment for age at diagnosis, sex, stage of cancer, BMI, diabetes duration, smoking status, HbA1c level, use of aspirin, use of insulin, sulfonylurea and thiazolidinediones, metformin was associated with a lower risk of overall mortality (HR, 0.66; 95% CI 0.476–0.923; p = 0.015) and CRC-specific mortality (HR, 0.66; 95% CI 0.45–0.975; p = 0.037) in CRC patients with diabetes (Table 3).
Table 2. Multivariate logistic regression analysis for CRC-specific mortality
Table 3. Univariate and multivariate adjusted overall mortality and CRC-specific mortality
We also performed a subgroup analysis based on the stage and location of cancer. While the CRC-specific mortality rate was not significantly different according to metformin use in stage 1, 2 and 4 CRC patients, stage 3 CRC patients exhibited a significantly higher CRC-specific survival rate (HR, 1.860; 95% CI 1.005–3.443; p = 0.048) and overall survival rate (HR, 1.722; 95% CI 1.010–2.935; p = 0.046) in the metformin group compared to the non-metformin group (Fig. 2). In our sub-analysis based on the cancer location, cancer-specific survival (HR, 1.740; 95% CI 1.008–3.005; p = 0.047) and overall survival (HR, 1.724; 95% CI 1.074–2.767; p = 0.024) in the metformin group were significantly higher than those in non-metformin group in patients with rectal cancer (Fig. 3). However, in the patients who had colon cancer, the cancer-specific mortality rate was not significantly different according to use of metformin.
In the present study, we found that CRC patients with diabetes who took metformin had lower mortality than those who did not take metformin. In addition, metformin use was an independent predictive factor of CRC-specific survival after adjustment for age at diagnosis, sex, stage of cancer, BMI, diabetes duration, smoking status, HbA1c level, use of aspirin and the use of insulin, sulfonylurea and thiazolidinediones. Previous studies have suggested that metformin decreases the incidence of cancer and cancer-related mortality in diabetic patients.6, 16, 17 However, these population studies lacked information pertaining to how the specific characteristics of cancers affected the metformin-related prognosis, namely, type, stage and treatment modality. Thus, our study is the first to show the positive effect of metformin on the mortality of CRC in diabetic patients.
Metformin has been widely used for a long time in the treatment of type 2 DM and is generally considered to be a safe and inexpensive drug. Several possible mechanisms for the anti-tumor effects of metformin have been proposed, although the mechanism most likely to be responsible for the majority of metformin's effects is through the inhibition of the mTOR signaling.19 Specifically, mTOR phosphorylates down-stream mediators regulating cell cycle progression, cell growth and angiogenesis. As the mTOR pathway is frequently deregulated in a variety of tumors, and activation of mTOR-dependent protein translation is associated with adverse prognosis and resistance to both chemotherapy and targeted therapy,20 inhibition of mTOR signaling is regarded as a promising cancer therapy target. In addition to inhibition of mTOR, other suggested anti-tumor mechanisms of metformin include decreased insulin growth factor (IGF)-I, blockage of HER-2 signaling, inhibition of angiogenesis, apoptosis and induction of cell cycle arrest.21–26
In our study population, insulin use and thiazolidinedione use were different between the metformin group and the non-metformin group. Importantly, insulin has been known to promote tumorigenesis, on which there has been recent controversy.6, 27 In addition, the evidence of potential anti-cancer effect of thiazolidinedione was reported in several studies.28, 29 However, when we adjusted for these variables in our analysis, use of metformin continued to be an independent factor of better outcome. Because the exact effect of insulin and thiazolidinedione on cancer was not proven and insulin (77) and thiazolidinedione (36) users were too small to evaluate the association with survival, insulin and thiazolidinedione use could not affect CRC-specific mortality and overall mortality. We also examined glycosylated hemoglobin levels and pre-meal glucose levels, which represent patient control of diabetes, to exclude the effects of improved diabetic control. In our study, glycosylated hemoglobin levels and pre-meal glucose levels were not different between the metformin and the non-metformin group, and therefore the anti-tumor effect of metformin was due to metformin itself rather than well-controlled diabetes.
Interestingly, we found that the effect of metformin on the survival rate of CRC was only present in a particular group of patients. Specifically, when we divided the patients according to tumor stage, the positive effect of metformin was not observed in patients with stage 1, 2 and 4 tumors. Indeed, only stage 3 patients had improved survival in the metformin group compared to those in the non-metformin group. However, the patient number of other groups was too small to evaluate the effect of metformin on survival. In our study, most of the stage 3 patients received adjuvant chemotherapy, which plays an important role in the clinical outcomes of stage 3 CRC. Therefore, we suggested that metformin has a potential synergic effect with other anti-cancer agents and might be a candidate drug as an additional therapy to adjuvant chemotherapy in locally advanced CRC. In one experimental study, metformin selectively targeted cancer stem cell and acted together with chemotherapy.30 To clarify this effect, we performed survival analysis about adjuvant chemotherapy regardless of stage of cancer and metformin use. Metformin use was associated with increased CRC-specific survival only in adjuvant chemotherapy group (Log Rank p = 0.019). In addition, when analyzed based on the location of cancer, the survival benefits of metformin use were present for rectal cancer, but not colon cancer. CRC exhibits different clinical characteristics and tumorigenic pathways according to the location of the tumor, including different molecular pathways, microsatellite stability and prognosis.31 Therefore, it is necessary to perform a more detailed examination of the molecular pathways targeted by metformin and their association with known tumorigenic pathway models.
There are several limitations in this study. First, we had inherent limitations associated with a retrospective study of non-randomly distributed patients. In addition, we had a limited sample size, especially in subgroup analysis based on stages and locations, and there was a potential risk of type I errors (false positive result). Second, in our study, consistent metformin use before and after CRC diagnosis was included. Therefore, it is unclear whether the observed benefit was from prediagnosis and/or postdiagnosis metformin use. To overcome these limitations, further studies such as large-scale prospective, observational studies and interventional trials as adjuncts to routine adjuvant chemotherapies are required. Third, we could not completely exclude the possibility that metformin use may be related to other prognostic factors, including other medications such as insulin and metabolic status. However, our findings remained unchanged after adjusting for these potential factors for CRC mortality, and there were no significant differences between the metformin and non-metformin groups with respect to BMI, plasma total cholesterol, pre-meal glucose and hemoglobin A1c levels, which represent the metabolic status and severity of diabetes.
In conclusion, CRC patients with diabetes receiving metformin had lower mortality than those not receiving metformin. To our knowledge, this is the first data to demonstrate an anti-tumor effect of metformin against CRC in diabetic patients. Metformin may have the potential to be useful in the adjuvant setting for locally advanced CRC. Further studies to evaluate the potential of metformin as an anti-tumor agent are warranted.