ASCO 2013

Effect of metformin on cancer risk and mortality

Dr Sara Gandini - European Institute of Oncology (IEO), Milan, Italy

We carried out a literature search, comparative literature search, looking at all studies, observational studies, and clinical trials on metformin and cancer risk and cancer mortality.

What did you examine?

We carried out a meta-analysis updating previous meta-analyses but at this time we focussed our analysis looking at all the possible bias that could confound our results and to have a more reliable estimate because most of them were observational studies so with many sources of bias. So, for example, we took into account if they were adjusted for BMI but also if there was a bias due to time-related bias, so immortal time bias. Actually there was a nice paper published by Swiss et al showing that many of the observational studies are false or the results are not so reliable due to this source of bias. So what we did was to stratify all our results looking at the studies that we were able to take into account with a proper analysis.

What were these studies looking at?

We looked at the association of metformin use, every user, and cancer incidence, cancer mortalities and in particular which kind of cancer. What we found was an overall significant association of metformin with the cancer incidence and cancer mortality but when we took into account the sources of bias the risk reduction was not so high so we got something like a 10% significant reduction in cancer risk and also a possible association with cancer mortality. In particular we found an association with breast and liver but still, again, the reduction was not so high, so probably 5% of significant reduction for breast cancer thanks to metformin use.

Is it not as good as we thought?

Yes, probably the reduction in risk is not so high but still it looks like there is an effect so it’s interesting anyway. Obviously the results were on diabetic people so were not in the general population, even if there were some studies looking also at the general population but mainly the results. So we need trials showing the effect on the general population, not only on diabetic people.

What could be the implications of this?

Possibly. We need to replicate the results with the trials because only the clinical randomised trial can really give us an answer because of the many sources of bias of observational studies. Then if we have an answer also in the general population it could be that maybe for subjects with a greater BMI it could be a possibility to think about.