Improving effiency of phase II oncology trials using RECIST endpoints
Session type: Oral
In phase II trials in solid tumours, endpoints are generally based on the RECIST criteria. This categorises patients based on their percentage change in tumour size and other reasons for non-response such as whether a patient develops new lesions. Normally these endpoints are analysed in a binary way – i.e. estimating the proportion of patients who are responders. However this ignores the information contained in the change in tumour size. Dichotomising continuous outcomes loses a lot of information and means higher sample sizes are needed for the same power. Analysing the continuous change in tumour size, although statistically efficient, is not always clinically meaningful as it ignores the other reasons for non-response.
In this talk I discuss a method to analyse the clinically relevant response endpoint, but using the information contained in the change in tumour size. This uses a statistical model of the different components of the responder based outcome. The method results in considerably higher power without any additional cost. Generally the increase in power is equivalent to increasing the sample size by around 35%. I will also discuss current work that aims to extend this method to progression-free survival.