Epidemiological studies are difficult to do well, pharmacoepidemiological studies are particularly complicated and studies of insulin and cancer risk are even more difficult to do well. “Well” in the sense that any association found can be interpreted in an unambiguous manner.

According to a report published today in ecancermedicalscience, the biggest hazard in pharmacoepidemiology is that with no randomised allocation possible, the probability of patients being allocated to particular treatments for un-identified reasons exists. In studies of insulin use and cancer risk, changes in different types of insulin prescribed and changes in dose of insulin create particular problems.

There has been controversy in these past months regarding insulin use and cancer risk, as reported in a previous report in ecancermedicalscience, and a variety of weaknesses in the epidemiological approach have been identified.

A group of international experts has been convened by the International Prevention Research Institute (iPRI) to review previous publications and existing resources in an attempt to provide guidance to the non-epidemiological reader on how to evaluate the strengths of the methodology used in such studies.

A check-list of essential items which should be present in epidemiological studies investigating insulin and cancer risk has been prepared and, according to iPRI President, Professor Peter Boyle, if there are a series of “no” responses then the findings of such a report should be treated with caution or dismissed entirely.

Professor Boyle added "the issue is not academic. There are approximately 200 million persons alive today with diabetes and an increasing proportion of them are using insulin. Even a small increase in risk of cancer with particular insulins would present an important Public Health problem. It is, therefore, essential that any such association reported is based on strong epidemiological methodology."