AI-powered Model Sharpens Treatment Decisions for Spinal Metastasis Patients

Using a machine learning technique known as Least Absolute Shrinkage and Selection Operator logistic regression, the researchers identified five preoperative factors that most strongly predicted one-year survival.

An AI-powered prognostic model developed by researchers at Nagoya University, Japan, has improved the accuracy of survival prediction for patients with spinal metastasis, offering clinicians a more reliable way to decide between surgery and palliative care in the era of advanced cancer therapies.

Spinal metastasis occurs when cancer spreads to the spine, affecting a large proportion of patients with advanced malignancies. Treatment choices typically depend on expected survival, with surgery considered for patients likely to live longer and palliative approaches favored when life expectancy is limited.

For decades, clinicians have relied on prognostic scoring systems developed using data from the 1990s and early 2000s, a period before targeted therapies and immunotherapy significantly extended survival for many cancer patients.

In a study published in the journal Spine, researchers from Nagoya University Graduate School of Medicine addressed this gap by developing a new AI-based model trained on prospective data from patients treated with modern oncologic approaches.

“Traditional survival prediction models in clinical practice use data from the 1990s and 2000s,” said Assistant Professor Sadayuki Ito, the study’s first author. “Those models don't fully reflect the impact of modern oncologic therapies, such as molecularly targeted therapies and immune checkpoint inhibitors.”

The research team conducted a large, multicenter prospective study involving patients who underwent surgery for spinal metastasis at hospitals across Japan. Using a machine learning technique known as Least Absolute Shrinkage and Selection Operator logistic regression, the researchers identified five preoperative factors that most strongly predicted one-year survival.

These included patient vitality as measured by the “Wake Up” component of a vitality index, age, functional status based on the ECOG scale, the presence of bone metastases outside the spine, and preoperative opioid use.

The AI-powered model achieved a strong predictive performance, with an area under the receiver operating characteristic curve of 0.762. Based on the scoring system, patients were grouped into low-, intermediate-, and high-risk categories, with one-year survival rates of 82.2 per cent, 67.2 per cent, and 34.2 per cent, respectively.

This level of stratification allows surgeons to better match treatment intensity with expected outcomes and tailor post-operative care more precisely. While the model has been developed using Japanese clinical data, the researchers see broader potential. “Our next step is to validate this system with data from medical institutions worldwide to ensure it can help patients globally,” concluded Dr Ito.

Stay tuned for more such updates on Digital Health News