Δαφνή, Ουρανία
Δεμίρης, Νικόλαος
Καρλής, Δημήτριος
Clinical trials have contributed in a great extent to the evolution of medicalresearch, through the examination of treatment efficacy. Tumor biology haschanged in a way that most tumors are heterogeneous and only a subset ofpatients with a particular cancer has the potential to benefit from a treatment. This phenomenon was created a shift towards the traditional way ofcomparing treatments. In recent years, most clinical trials are conductedin a 2-stage design, where first a sensitive subgroup is identified and thenthe treatment efficacy is calculated for both the sensitive subgroup and allpopulation.In this study we are interested in creating an adaptive design for detecting gene signatures and prospectively identify a target subpopulation whenhaving time-to-event endpoints as a primary outcome. For this purpose, wewill use the Cox proportional hazard model with only parameters the treatment effect and its interaction with the expression genes. This interactionhas a crucial role in our design, as it determines which genes will be includedin our signature. Moreover, as far as the signature is developed, we will seta classification rule based in the genes belonging in the signature and thehazard ratio of each patient, to identify the sensitive subpopulation.Finally, we are focusing to the evaluation of the treatment by conductingone hypothesis test for all randomized patients and a second only for the sensitive subgroup. The performance of the design will be measured in terms ofstatistical power, regarding two different scenarios and by alternating severalparameters such as the sample size.
Εκφράσεις γονιδίων
Σχεδιασμός
Ευαίσθητη υποομάδα
Καρκίνος
Clinical trials
Gene expressions
Adaptive design
Sensitive subgroup
Cancer