Advanced methods in clinical trial design and analysis

The present dissertation is based on a collection of papers motivated by the need to improve the methodological foundations of clinical research in an era where both randomised clinical trials (RCTs) and real-world data (RWD) play central, complementary roles in evidence generation. While RCTs remain the gold standard for evaluating efficacy and safety, their rigidity, high cost, and limited external validity highlight the importance of more flexible and efficient trial designs. At the same time, the growing availability of routinely collected healthcare data offers unprecedented opportunities for causal inference and predictive modelling, but also raises methodological challenges that demand innovative statistical solutions.The thesis is structured into two main parts, each addressing one of these domains. Section I focuses on methodological contributions to the design of clinical trials. It investigates modern adaptive approaches, with particular attention to Bayesian dynamic borrowing within Group-Sequential Designs for medical devices studies and safety-driven response-adaptive randomisation in oncology trials. These contributions aim to enhance efficiency, ethical oversight, and clinical relevance, while ensuring statistical rigour and regulatory acceptability. Section II turns to real-world data, which provide a valuable complement to experimental evidence. This part discuss methods for causal inference using win-based statistics in observational oncology studies with composite endpoints and explores predictive modelling strategies in paediatric intensive care. Emphasis is placed on addressing challenges such as confounding, heterogeneity, imbalance, and low eventrate outcomes. Together, the two parts form a coherent programme of research that bridges statistical innovation with pressing needs in clinical and regulatory science.