Abstract Identifying appropriate preclinical cancer models remains a major challenge in increasing the efficiency of drug development. A potential strategy to improve patient outcomes could be selecting the ‘right’ treatment in preclinical studies performed in patient-derived xenografts (PDXs) obtained by direct implants of surgically resected tumours in mice. These models maintain morphological similarities and recapitulate molecular profiling of the original tumours, thus representing a useful tool in evaluating anticancer drug response. In this review, we will present the state-of-art use of PDXs as a reliable strategy to predict clinical findings. The main advantages and limitations will also be discussed.