Guidelines for investigation of associations between insulin and cancer risk using prescription databases.

e12061 Background: Analysis of associations between insulin and cancer risk using Prescription Databases and Cancer Registers is complicated and should only be undertaken with extreme caution and employing advanced statistical and epidemiological methodologies. Methods: Interaction between methodologists, diabetologists and oncologists is also critical. Failure to do so can result in unfortunate events with potential societal implications as could have happened following the recent publication of a flawed database analysis of the association between a long-acting insulin analogue and an increased risk of cancer. Such reports cause unnecessary confusion and concern among patients and their physicians. Results: Prescription Databases were not established with such analyses in mind, and cannot identify many critical limitations such as changes in insulin type during study observation (“switchers”). Notably, those who use more than one type of insulin simultaneously pose specific problems but restriction of the analysis to users of only one type of insulin during the study period is inappropriate and likely to introduce bias. There are outstanding issues surrounding how to deal with the entire issue of “dose” which requires further exploration by the biostatistical community (reliability of doses captured, defining doses at baseline or/and differences in dose exposure). Conclusions: It is possible to obtain clear information from Prescription Databases if (and only if) the study is designed adequately, key data are available including multiple confounders (age, body weight, diabetes duration glucose control, socioeconomic status, etc.) and the biostatistical analyses are predefined. Author Disclosure Employment or Leadership Position Consultant or Advisory Role Stock Ownership Honoraria Research Funding Expert Testimony Other Remuneration Novo Nordisk, sanofi-aventis Novo Nordisk, sanofi-aventis sanofi-aventis