An international multi-centre study to develop and validate federated learning-based prognostic models for anal cancer - PubMed
5 hours ago
- #federated learning
- #anal cancer
- #prognostic models
- International multi-centre study develops federated learning-based prognostic models for anal cancer.
- Federated learning used to train models across 14 centres (1428 patients) and validated in 2 additional centres (277 patients).
- Models achieved consistent calibration and discrimination (c-indices 0.68-0.79).
- Key factors for improved overall survival: lower T stage, absence of nodal involvement, smaller tumour volume, female sex, younger age, and specific chemotherapy types.
- Improved locoregional control linked to lower T stage, smaller tumour volume, and female sex.
- Better freedom from distant metastases associated with absence of nodal involvement and smaller tumour volume.
- Demonstrates federated learning enables privacy-preserving prognostic modelling for rare cancers without data sharing.
- Leonard Wee and Andre Dekker have competing interests; other authors declare none.