Routine laboratory trajectories encode the onset of organ-level complications in cancer

2026-06-07

AI Summary

Current single-timepoint prognostic models struggle to capture early signals of organ-level complications during cancer therapy. This study proposes a Transformer-based temporal modeling approach that leverages longitudinal routine laboratory data from nearly 4,000 patients with multiple myeloma or ovarian cancer to predict 162 treatment-related complications over the subsequent two years. We demonstrate for the first time that trajectories of routine laboratory values can identify organ dysfunction weeks to months in advance, enabling complication-specific monitoring without additional testing. The model’s generalizability is validated across diverse datasets, including MIMIC-IV and MMRF CoMMpass, spanning different cancer types and healthcare systems. It achieves 1.5–6.1-fold enrichment in prevalence across eight complication categories, with AUROC scores up to 0.85—significantly outperforming non-sequential baselines by as much as 0.11.

Abstract

Routine laboratory panels drawn during cancer treatment constitute longitudinal physiological recordings of organ function, yet their temporal structure is discarded by single-timepoint prognostic tools. A transformer trained on 2,777,595 laboratory measurements from 3,905 patients with multiple myeloma or ovarian cancer predicted the two-year onset of 162 treatment-associated complications, including therapy-related myelodysplastic syndromes, spanning eight clinical categories, achieving 1.5- to 6.1-fold enrichment above prevalence at the group level. It matched or outperformed non-sequential baselines across grouped endpoints (AUROC gains up to +0.11), demonstrating that longitudinal laboratory trajectories capture evolving complication-specific physiology inaccessible from isolated measurements. Predictions generalised across both cancers, divergence concentrating in disease-specific complications, and biomarker masking recovered signatures consistent with established pathophysiology. External validation on MIMIC-IV and MMRF CoMMpass confirmed transferability across independent healthcare systems (AUROC up to 0.85). Routine oncological laboratory data encode organ deterioration weeks to months before clinical onset, enabling complication-specific surveillance without additional testing infrastructure.

Problem

• longitudinal laboratory trajectories
• treatment-associated complications
• organ dysfunction
• cancer therapy
• early prediction

Innovation

• longitudinal laboratory trajectories
• transformer model
• treatment-associated complications
• early prediction
• organ-level deterioration