The post hoc use of randomised controlled trials to explore drug associated cancer outcomes: methodological...

Drug-induced cancer risk is of increasing interest. Both observational studies and data from clinical trials have linked several widely used treatments to cancer. When a signal for a potential drug-cancer association is generated, substantiation is required to assess the impact on public health before proper regulatory action can be taken. This paper aims to discuss challenges of exploring drug-associated cancer outcomes by post-hoc analyses of Randomised controlled trials (RCTs) designed for other purposes. METHODOLOGICAL CHALLENGES TO CONSIDER: We set out to perform a post-hoc nested case-control analysis in the ADVANCE trial in order to examine the association between insulin use and cancer. We encountered several methodological challenges that made the results difficult to interpret, including short duration of exposure of interest, lack of power, and correlation between exposure and potential confounders. Considering these challenges, we concluded that using the data would not enlighten the discussion about insulin use and cancer risk and only serve to further complicate any understanding. Therefore, we decided to use our experience to illustrate methodological challenges, which need to be addressed when re-analysing trial data for cancer related outcomes.

Substantial amount of information on cancer outcomes is available from RCTs. Hence, making use of such data could save time and spare patients from inclusion in further trials. However, methodological challenges must be addressed to enhance the likelihood of reliable conclusions. Advantages of post-hoc analyses of RCTs include quality of data collected and sometimes randomisation to exposure of interest. Limitations include confounding and sample size, which is fixed to suit the purposes of the trial, insufficient duration of exposure and identification of underlying biological mechanisms relating treatment to cancer to formulate the most appropriate post-hoc study design.