Predictive Process Monitoring (PPM) is a subfield of Process Mining, which focuses on using Machine Learning (ML) methods to predict the future of an ongoing execution based on its early stages. Predictions provided by PPM approaches include the time remaining until the end of the process execution, the next event activity to be performed, or the overall outcome of the execution trace. While different machine and deep learning methods have been leveraged in the literature of PPM, in recent years, Graph Neural Networks (GNNs) have emerged as a new type of neural network and have been leveraged by few approaches in PPM. GNN models offer the advantage of working with a more natural representation of a trace prefix, allowing more expressive and semantically rich encodings. In this paper, we review three GNNs-based approaches from the PPM literature and we compare them with state-of-the-art approaches on the task of next activity prediction. The results show that GNNs are able to achieve an accuracy gain of more than 10% with respect to traditional approaches for some datasets, thus making them a promising solution in PPM.