LIMBS-PAD: A Synthetic Multimodal Longitudinal Dataset for Peripheral Artery Disease

LIMBS-PAD is a fully synthetic longitudinal cohort centred on peripheral artery disease (PAD), built for methods development rather than epidemiological description. No real individuals are represented and the data describe no real population, and there are no consent or disclosure-risk constraints on reuse. The readme markdown document provides context and a human-readable explanation of the deposit, which provides a 500-participant sample: 2,186 attended visit rows across five annual visits, as six relational tables in both CSV and Parquet, with a 942-variable data dictionary and a Croissant machine-readable description. Modalities span clinical measurement, PAD phenotyping, 502 protein assays, 248 NMR metabolites, physical function and socioeconomic position. The accompanying Study Specification and Cohort Profile describe the complete 20,000-participant build (88,525 visit rows), of which the data here are a sample. The design is enriched: PAD prevalence exceeds the 3-7% in this age band. The panel is unbalanced through attrition, and the dictionary flags outcome-leaking variables, supporting multimodal and longitudinal methods work. This dataset is intended as a descriptor for the larger dataset, to allow for hypothesis generation based on data types and quantities.