Constructing estimands for overall survival within causal frame work for dynamic treatment schemes in oncology: Simulation and a clinical trial example

Overall survival (OS) is an important endpoint in regulatory approval and health technology assessment (HTA). Using current statistical methods, overall survival rates for patients given new therapies in clinical trials may be overestimated if a non-trivial proportion of the patients receive treatment changes after the trial. Adjustments need to be made to cancer trials to take into account the availability of new treatments which may skew the survival rates. Some statistical methods have been proposed in recent research to adjust the biased overall survival rate of clinical trial participants, but there are issues with these methods as, for example, although some disease-related patient characteristics were considered in these methods, they fail to recognize that the patient characteristics could change over time. Moreover, the current statistical methods assume that all the treatments a patient may receive for the disease after a clinical trial will have the same effect on that disease, whereas a new innovative treatment may have a much greater impact. Hence, in this project we are introducing a new statistical method to estimate overall survival for patients who receive a subsequent anticancer treatment, where the bias can be controlled. Our aim is to develop a new method for analyzing clinical trial data to give more realistic overall survival rates.