JMIR Publications

ABSTRACT

Background:

Patients with bone metastasis often experience a significantly limited survival time, and a life expectancy of less than three months is generally regarded as a contraindication for extensive invasive surgeries. In this context, the accurate prediction of survival becomes of utmost importance as it serves as a crucial guide in making clinical decisions.

Objective:

The objective of this study was to develop a machine learning-based prediction model that can provide a more accurate assessment of the likelihood of early death among patients with bone metastasis.

Methods:

This study conducted an analysis of a large cohort consisting of 118,227 patients diagnosed with bone metastasis between 2010 and 2019, using data obtained from a national cancer database. The occurrence of early death among these patients was defined as those who had passed away either at or within three months after the diagnosis of bone metastases, with the cause of death attributed to cancer-related factors. The entire cohort of patients was randomly split 9:1 into a training group (n=106492) and a validation group (n=11735). Six approaches including logistic regression, extreme gradient boosting (XGBoosting) machine, decision tree, random forest, neural network, and gradient boosting machine were implemented in the study. The performance of these approaches was evaluated using eleven measures, and each approach was ranked based on its performance in each measure. The optimal model was determined as the approach with the highest cumulative score across all measures. Patients (n=332) from a teaching hospital was regarded as the external validation group, and external validation was performed using the optimal model.

Results:

In the entire cohort, a significant proportion of patients (36.6%) experienced early death. Multivariate analysis revealed several factors that were significantly associated with early death, including age, sex, race, marital status, rural-urban continuum, primary site, tumor stage (T stage), node stage (N stage), brain metastasis, liver metastasis, lung metastasis, cancer-directed surgery, radiation, and chemotherapy. These factors were considered for model training and optimization. Among the different approaches evaluated, the gradient boosting machine exhibited the highest score of prediction performance (54 points), followed by the neural network (52 points) and XGBoosting machine (50 points). The gradient boosting machine demonstrated a favorable discrimination ability, with an area under the curve (AUC) of 0.858 (95% CI: 0.851-0.865). Additionally, the calibration slope was 1.02, and the intercept-in-large was -0.02, indicating good calibration of the model. Risk stratification of patients was achieved based on the developed model using the gradient boosting machine. Patients were divided into two risk groups using a threshold of 37.00%. Patients in the high-risk group (71.96%) were found to be 4.5 times more likely to experience early death compared to those in the low-risk group (15.62%). External validation of the model demonstrated a high AUC of 0.847 (95% CI: 0.798-0.895), indicating its robust performance. The calibration slope was 1.06, and the intercept-in-large was 0.17, further supporting the validity and reliability of the model.

Conclusions:

This study develops a machine learning-based prediction model to accurately assess the probability of early death among bone metastasis patients. The gradient boosting machine demonstrates the highest prediction performance among the six approaches evaluated. This prediction model has the potential to guide clinical decision-making and improve the care of bone metastasis patients by identifying those at a higher risk of early death.