[DATASET] GAMA-PCa: 548 synthetic Spanish prostate cancer cases

GAMA-PCa is a dataset of 548 fully synthetic longitudinal prostate cancer cases in Spanish. Each case includes both a clinical narrative and its corresponding structured JSON case dossier (“skeleton”), describing patient characteristics, diagnostic findings, risk classification, initial treatment and longitudinal clinical events. These paired representations support research on clinical information extraction, structured-to-text generation and consistency between structured data and clinical narratives. The corpus covers 16 predefined clinical archetypes, with 34–35 cases per archetype. It includes silver-standard entity annotations in BRAT format, model-specific annotation outputs and LLM-as-a-judge evaluation reports. Historical narratives superseded by revised versions are explicitly marked as non-final and are not counted as additional cases. A 48-case subset underwent independent clinical evaluation by three clinical reviewers. This assessment does not imply expert validation of all 548 cases or of the silver-standard annotations. All patients are fictional. The dataset is intended for research, not clinical decision-making, and is distributed under the Creative Commons Attribution-NonCommercial 4.0 International license (CC BY-NC 4.0).