Nick Mulcahy

May 13, 2015

PHILADELPHIA — The great hope that the diabetes drug metformin will be a life-extending treatment for pancreatic cancer looks a bit shaky, according to a new study.

Hope has been rooted in a couple of factors, suggested lead author Roongruedee Chaiteerakij, MD, PhD, from the Mayo Clinic Cancer Center in Rochester, Minnesota.

First, there is the profound need for new treatment. Pancreatic cancer has a median survival of only 5 to 8 months, and an overall 5-year survival rate of only 6%, she said.

Second, multiple epidemiologic studies have suggested there is a reduction in cancer mortality, including pancreatic cancer mortality, with metformin.

This has led to a flurry of at least 20 investigations of metformin in a variety of cancers. In fact, there are currently 10 clinical trials on the use of metformin in the treatment of pancreatic cancer that are ongoing or that have not finalized and reported their data, Dr Chaiteerakij told reporters during a press conference here at the American Association for Cancer Research 2015 Annual Meeting.

However, the original epidemiologic evidence that inspired those trials might have been flawed, she explained. In fact, results from a study by she and her colleagues serve as a "cautionary lesson" about the shortcomings of standard epidemiologic data.

In patient databases, metformin exposure is "commonly characterized as simply" ever or never, she pointed out. This could "potentially introduce unintended biases," she said. Most notably, there could be patient-selection bias because ever users might be, collectively, healthier and living longer than those who never got a prescription for the drug.

Dr Chaiteerakij and her colleagues investigated the matter with a retrospective cohort study of 980 patients with pancreatic ductal adenocarcinomas (PDAC) and diabetes at the Mayo Clinic.

Detailed patient charts allowed them to look at metformin use in PDAC patients in a more exact way than the typical patient database, she said.

First, the team ran a conventional analysis of the data that looked at the standard ever or never use of metformin.

Then they ran a more detailed analysis. This time, they compared never users with four groups of ever users, classified according to initiation of metformin: initiation more than 1 year before PDAC diagnosis, initiation less than 1 year before PDAC diagnosis, initiation less than 30 days after PDAC diagnosis, and initiation more than 30 days after PDAC diagnosis.

Refining the statistical analysis made a difference, Dr Chaiteerakij reported.

In the conventional analysis (ever or never), metformin use was associated with a "marginally significant" improvment in survival of 1 month in PDAC patients. In other words, it suggested that metformin might improve survival a bit.

In the refined analysis, time between diagnosis and initiation of metformin was associated with survival. Specifically, median survival was significantly longer in patients who started the drug more 30 days after diagnosis than in never users. This group was the only one with a significant increase in survival.

Table. Hazard Ratios of Ever Users Compared With Never Users

Metformin Initiation Relative to PDAC Diagnosis Hazard Ratio 95% Confidence Interval
>1 year before 1.08 0.85–1.37
<1 year before 0.99 0.85–1.17
<30 days after 1.04 0.83–1.31
>30 days after 0.49 0.33–0.74


"These patients already survived more than 30 days, suggesting that there is an inherent survival bias in this group of patients," Dr Chaiteerakij said. "After accounting for these unintended biases, the benefit of metformin was not confirmed in our study."

"Patients starting metformin after PDAC diagnosis lived long enough to use metformin," she added.

The data "collected do sort of qualify the enthusiasm" for the other clinical trials of metformin in pancreatic cancer, said William Nelson, MD, from the Sidney Kimmel Comprehensive Cancer Center at Johns Hopkins University in Baltimore, who moderated the press conference during which Dr Chaiteerakij spoke.

"This study highlights the importance of appropriate design of retrospective studies and the necessity of conducting prospective studies with solid rationale for determining the effect of diabetes drugs on cancer risk or death," said Dr Chaiteerakij in a press statement.

This study was supported in part by a National Cancer Institute SPORE research grant. Dr Chaiteerakij has disclosed no relevant financial relationships.

American Association for Cancer Research (AACR) 2015 Annual Meeting: Abstract LB-183. Presented April 21, 2015.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.