Diabetes Mellitus and Increased Risk of Cancer: Focus on Metformin and the Insulin Analogs
M. Shawn McFarland, Pharm.D.; Rebecca Cripps, Pharm.D.
Posted: 11/18/2010; Pharmacotherapy. 2010;30(11):1159-1178. © 2010 Pharmacotherapy Publications
Abstract and Introduction
Type 2 diabetes mellitus has been associated with an increased risk of hepatic, pancreatic, colon, endometrial, breast, and bladder cancer. Although a mechanism of action for the increased risk has been postulated, no definitive evidence has been completely elucidated in the medical literature. Results of recently released studies documented the use of specific antidiabetic drugs with increased rates of cancer. The insulin analog glargine was the focus of four observational studies published in 2009 that outlined an increase in the rates of cancer associated with its use. In contrast, the use of metformin has been shown to possibly decrease the rate of specific cancers when used in the treatment of type 2 diabetes. These data regarding cancer risk and antidiabetic drugs are contradictory and at this time are inconclusive. Until results of long-term randomized prospective studies are available to elucidate a correlation with cancer and insulin, we must continue treating diabetes in order to avert the long-term complications of the disease.
Diabetes and Cancer
Risk of Cancer
Epidemiologic studies have linked diabetes to different types of cancer.[17–27] Dating back almost 20 years, meta-analyses have attempted to review the literature with regard to the incidence and mortality rates of specific cancers in patients with diabetes. Specifically, higher rates of pancreatic and hepatic cancers have been identified.
Pancreatic Cancer In one meta-analysis, the risk of pancreatic cancer and type 2 diabetes was evaluated. By searching multiple biomedical databases, the authors identified 36 studies from 1973–2005. Age- and sex-adjusted odds ratios (ORs) for pancreatic cancer associated with type 2 diabetes were extracted. Seventeen were case-control studies and 19 were cohort or nested case-control studies. Information was available on a total of 9220 patients. For the 17 case-control studies, a significant association was noted between type 2 diabetes and pancreatic cancer (OR 1.94, 95% confidence interval [CI] 1.43–2.46). Cohort studies yielded a similar finding, with an OR of 1.73 (95% CI 1.59–1.88). Combined OR for all studies was 1.82 (95% CI 1.66–1.99). The authors noted that there was some heterogeneity among both study designs (p=0.002 for case-control studies, p=0.05 for cohort studies) that was not explained by differences in sex, adjustment for smoking, or the method of diagnosing diabetes. The explanation given was the higher risk of pancreatic cancer reported in smaller studies and studies performed before the year 2000. Of the 36 studies, nine reported categories regarding disease duration. Evaluation of these studies revealed a 50% greater risk of developing pancreatic cancer in individuals with disease duration of less than 4 years versus individuals who had diabetes for more than 4 years (OR 2.1, 95% CI 1.9–2.3 vs OR 1.5, 95% CI 1.3–1.8). The authors concluded that along with cigarette smoking and obesity, type 2 diabetes may be a likely modifiable risk factor for pancreatic cancer.
Hepatocellular Cancer An evaluation of the literature regarding diabetes and hepatocellular carcinoma was conducted in 2006. The authors performed a MEDLINE search for published articles from January 1966–February 2005 that pertained to diabetes and hepatocellular carcinoma. Twenty-six studies met the inclusion criteria (13 casecontrol studies and 13 cohort studies). Of the 13 case-control studies, eight found a statistically significant positive association between diabetes and hepatocellular carcinoma. The pooled OR for all case-control studies was 1.84 (95% CI 1.25–2.69). Confounders were adjusted for in eight of the 13 studies. Specifically, adjustment for alcohol use and viral hepatitis did not change the positive association between diabetes and hepatocellular carcinoma. In these 13 studies, diabetes was defined by face-to-face or mailed questionnaires in 10 studies and by medical record review or diagnostic codes in three studies.
Of the 13 cohort studies, seven evaluated patients with diabetes specifically, three studied a general population cohort, and three studied patients with underlying chronic liver disease. Diabetes was diagnosed based on glucose levels in five studies, use of a prescription drug for diabetes in two studies, diagnostic codes in three studies, and questionnaires in three studies. The pooled OR for all cohort studies was 2.5 (95% CI 1.93–3.24). Among the seven cohort studies of persons with diabetes, the pooled OR was 2.50 (95% CI 1.72–3.61). Three of the seven studies found a statistically significant positive association between hepatocellular cancer and diabetes. Only three of the seven studies adjusted for confounders, and although the risk was weakened, a risk still was associated with diabetes and hepatocellular carcinoma.
For both case-control and cohort study designs, heterogeneity did not exist among studies that used population-based controls (p=0.26) but did exist among those with hospital-based controls (p<0.01). Also noted was that heterogeneity existed among U.S. and European studies, but not among Asian studies. The authors concluded that there was a 2.5-fold increase in the risk for hepatocellular carcinoma in patients with diabetes.
Colorectal Cancer A meta-analysis was conducted of published data on the association between diabetes and the incidence and mortality of colorectal cancer. The authors searched MEDLINE for data published from 1966–2005 with specific MeSH terms for the subject of interest. They identified 15 studies overall (6 case-control and 9 cohort studies) that met their predefined inclusion and exclusion criteria. Eight of the 15 studies found a statistically significant positive association between diabetes and colorectal cancer (average relative risk (RR) for all studies 1.30, 95% CI 1.20–1.40); no heterogeneity was found among these studies (p=0.21). Results did not differ depending on geographic location, as the results were significant whether the trial was conducted in Europe or the United States. The association did not differ with respect to sex. When restricted to studies controlling for physical activity and body mass index (BMI), two known confounders of the positive association of diabetes and cancer, a positive association still existed (summary RR 1.34, 95% CI 1.20–1.49).
Among the six cohort studies that evaluated mortality, a statistically significant positive association was noted between diabetes and colorectal cancer (summary RR 1.26. 95% CI 1.05–1.50), and significant heterogeneity existed among these studies (p=0.04). Excluding the study contributing to heterogeneity identified by a sensitivity analysis, the association between diabetes and mortality was not as strong (summary RR 1.9, 95% CI 1.10–1.28), with no statistically significant heterogeneity (p=0.40). Once again, results did not differ based on sex. The authors concluded that published data trended toward a positive association between colorectal cancer and diabetes.
Bladder Cancer In 2006, a meta-analysis was conducted that evaluated the medical literature for the risk of bladder cancer in patients with diabetes. The authors searched MEDLINE for data published from 1966–2006 with specific MeSH terms for the subject of interest. They identified 16 studies overall (seven case-control studies, three cohort studies, and six cohort studies of hospitalized patients with diabetes) that used external population comparisons and met their predefined inclusion and exclusion criteria. Evaluation of all studies revealed that diabetes was associated with a statistically significant increased risk of bladder cancer compared with patients without diabetes (RR 1.24, 95% CI 1.08–1.42). A strong indication of heterogeneity existed among the studies (p<0.0001).
When evaluated based on study design, there was a 40% increase in the risk of bladder cancer among cohort studies (summary RR 1.43, 95% CI 1.04–1.80) and case-control studies (summary RR 1.37, 95% CI, 1.04–1.80). However, there was no increased risk in cohort studies of patients with diabetes specifically (summary RR 1.01, 95% CI 0.90–1.12). Of interest, the summary estimates were significantly higher in studies conducted in North America than in Europe (p=0.07), for studies published after 2000 (p=0.01), and for studies that adjusted for smoking (p<0.0001). Summary estimates were similar in studies that adjusted for BMI. The authors concluded that patients with diabetes mellitus may have an increased risk for bladder cancer.
Breast Cancer In another study, the risk of and mortality from breast cancer and diabetes mellitus were evaluated.  The MEDLINE database was searched for articles related to the topic from 1966–2007. Twenty studies met the authors' predefined inclusion and exclusion criteria (5 case-control and 15 cohort studies). Nine trials were conducted in the United States, seven in Europe, and four in Asia. An increased risk of breast cancer was found in 15 studies, with the risk reaching statistical significance in eight. The RR was 1.18 (95% CI 1.05–1.32) for case-control studies and 1.20 (95% CI 1.11–1.30) for cohort studies. Among all studies, the RR was 1.20 (95% CI 1.12–1.28) for women with diabetes versus women without diabetes, indicating a 20% increased risk of breast cancer. Statistically significant heterogeneity existed among the studies (p=0.01). Using a sensitivity analysis, the authors excluded three studies that contributed significantly to the summary estimate, due to a diagnosis of diabetes that was based on discharge data. The RR remained unchanged but the CI widened and heterogeneity among the studies was reduced (RR 1.20, 95% CI 1.09–1.33, p=0.10).
Five of the studies reported mortality results on diabetes and breast cancer. Combined results of these studies provided a nonsignificant RR of 1.24 (95% CI 0.95–1.62). The authors concluded that diabetes, largely type 2, was associated with an increased risk of breast cancer but not an increase in mortality.
Endometrial Cancer Another group evaluated the risk of endometrial cancer and diabetes. A MEDLINE database search was performed for articles related to the topic from 1966–2007. Twenty-five studies met the authors' predefined inclusion and exclusion criteria: 13 case-control studies and 12 cohort studies (5 reported incidence and/or mortality rate ratios as the measure of RR, and 7 reported standardized incidence or mortality rates as the measure of risk). Twelve trials were conducted in the United States, 11 in Europe, one in Asia, and one in South America. In this meta-analysis of endometrial cancer and diabetes (largely type 2), 16 of the 25 studies were used (3 cohort and 13 case-control). Twelve of the 16 studies found an increased risk of endometrial cancer and diabetes with a summary RR of 2.10 (95% CI 1.75–2.53). Heterogeneity existed among the studies (p=0.01); however, when a sensitivity analysis was performed and individual studies were excluded one at a time, a significant positive association still remained (summary RR 1.98–2.18, with a 95% CI lower limit that never crossed 1.0). The association of endometrial cancer was somewhat stronger in studies conducted in Europe (summary RR 2.51, 95% CI 1.83–3.45) versus the United States (summary RR 1.70, 95% CI 1.47–1.98).
When the meta-analysis was restricted to the two studies that addressed the two confounders of physical activity and BMI, a positive association between endometrial cancer and diabetes still existed (summary RR 2.47, 95% CI 1.37–4.45). Of the two cohort studies that evaluated mortality, one reported a statistically significant increase and one did not. When the authors pooled the studies, a nonsignificant positive association existed between diabetes and mortality for endometrial cancer (summary RR 1.58, 95% CI 0.94–2.66). Unlike the previous meta-analyses, two cohorts provided standardized incidence ratios whereas one case-control study evaluated the association specifically in type 1 diabetes. When the three studies were combined, a positive association existed between type 1 diabetes and endometrial cancer (summary RR 3.15, 95% CI 1.07–9.29) with no evidence of heterogeneity (p=0.04). The authors concluded that there was a positive association between diabetes and endometrial cancer.
Non-Hodgkin's Lymphoma The risk of non-Hodgkin's lymphoma was evaluated in two meta-analyses.[30, 31] The first analysis was published in September 2008. The authors searched the MEDLINE and EMBASE databases for studies published before 2007 that reported an association between non-Hodgkin's lymphoma and diabetes. A total of 10 case-control and three cohort studies met inclusion criteria. Six were conducted in North America, four in Europe, and three in Asia. The meta-analysis for all studies combined showed a positive association between diabetes history and non-Hodgkin's lymphoma (RR 1.28, 95% CI 1.07–1.53). When analyzed individually by study design, a significant positive association was noted in hospital-based case-control and cohort studies (RR 1.36, 95% CI 1.00–1.86 and RR 1.79, 95% CI 1.30–2.47, respectively) and not in population-based case-control studies (RR 1.12, 95% CI 0.89–1.40). Significant heterogeneity existed in all included studies (p=0.05). In the studies that evaluated sex, a positive association was seen in women specifically (RR 1.60, 95% CI 1.15–2.22) versus men (RR 1.15, 95% 0.84–1.59). A significant difference was noted based on location, with a significant association in studies conducted in Europe (RR 1.23, 95% CI 0.99–1.52) and Asia (RR 1.74, 95% CI 1.31–2.33) versus North America (RR 1.11, 95% CI 0.82–1.50). When studies adjusting for the confounder of BMI were evaluated, there was still a statistically significant positive association (RR 1.56, 95% CI 1.23–2.00). The authors concluded that although a risk of non-Hodgkin's lymphoma seems to be present in patients with diabetes, their results were overall inconclusive.
Later in 2008, another group published results of their MEDLINE search for observation cohort and case-control studies from 1980–2008 that reported an association between non-Hodgkin's lymphoma and diabetes. Fifteen studies met the authors' predefined inclusion and exclusion criteria: 10 case-control studies and 5 cohort studies (3 reported incidence and/or mortality rate ratios as the measure of RR, and 2 reported standardized incidence or mortality rates as the measure of risk). All studies combined provided an RR of 1.19 (95% CI 1.04–1.35) for non-Hodgkin's lymphoma in patients with diabetes. When specific study types were evaluated, a statistically significant risk was seen in prospective studies (RR 1.41, 95% CI 1.07–1.88) with no heterogeneity (p>0.10) but not seen among case-control studies (RR 1.12, 95% CI 0.95–1.31) with some heterogeneity (p=0.09). In studies that reported risk according to sex, a statistically significant increased risk was apparent in women (RR 1.38, 95% CI 1.06–1.80) but not men (RR 0.98, 95% CI 0.79–1.22).
As in previous analyses, several studies used age of onset or used a national registry to evaluate the type of diabetes. The authors pooled the results of seven trials that used age of onset as a diagnosis of diabetes (> 30 yrs for type 2 and < 30 yrs for type 1). The pooled RR among patients with type 1 diabetes was 1.27 (95% CI 0.82–1.99) without evidence of heterogeneity (p>0.10), and the RR was 1.32 (95% CI 0.93–1.86) with heterogeneity (p=0.04) in patients with type 2 diabetes. The authors concluded that overall there seems to be a moderate risk for non-Hodgkin's lymphoma in patients with diabetes.
Summary Limitations exist among all meta-analyses. In most of the evaluated meta-analyses, individual studies did not distinguish between type 1 and type 2 diabetes, making it difficult to truly interpret the risk as a whole. In studies that defined diabetes, the diagnosis was made based on age, diabetes registry, or self-reported status. Evaluation of the risk of specific cancer with regard to diabetes can be confounded by multiple factors. Two of the most important would be obesity and physical inactivity. However, in the meta-analyses where adjustments for BMI and physical activity were made, a link still was present between diabetes and cancer. In all the evaluated meta-analysis publications, bias was possible. Finally, use of other drugs such as nonsteroidal inflammatory drugs or aspirin, which can modify overall cancer risk, may have been a confounder complicating interpretation of results from meta-analyses. Taking into consideration the published data, a link between diabetes and specific cancer types does seem to be probable. A review of the overall data is provided in Table 1.[2, 4–7, 28–31]