⚡ Quick Summary
This study developed three predictive models (PMs) to estimate overall survival (OS) for patients with bone metastases referred to palliative radiotherapy (RT). The models demonstrated promising performance with AUC values of 0.901, 0.767, and 0.806 for short-, intermediate-, and long-term OS, respectively.
🔍 Key Details
- 📊 Dataset: 567 patients from the PRAIS trial
- 🧩 Features used: Clinical and laboratory parameters
- ⚙️ Technology: Machine learning techniques, including LASSO
- 🏆 Performance: AUC values: 0.901 (3 weeks), 0.767 (24 weeks), 0.806 (52 weeks)
🔑 Key Takeaways
- 📊 Predictive models can significantly aid in assessing survival expectancy for patients with bone metastases.
- 💡 Clinical parameters such as primary tumor site and presence of non-bone metastases were crucial for OS prediction.
- 👩🔬 Laboratory parameters like interleukin 8 (IL-8) and C-reactive protein also played a significant role.
- 🏆 The models achieved high AUC values, indicating strong predictive capabilities.
- 🌍 This research is part of the PRAIS trial, a multicenter study focused on palliative care.
- 🔄 Further validation is needed before clinical implementation of these predictive tools.
📚 Background
Patients with bone metastases often experience debilitating pain, necessitating effective palliative care strategies. The decision to administer palliative radiotherapy (RT) hinges on accurately assessing the patient’s survival expectancy. This study aims to enhance decision-making processes by developing predictive models that utilize easily accessible clinical and laboratory data.
🗒️ Study
This research is a secondary analysis of the PRAIS trial, a longitudinal observational study that collected data from patients referred for palliative RT due to cancer-induced bone pain. The analysis included 567 patients, focusing on the correlation between various clinical and laboratory parameters and overall survival at three time points: short-term (3 weeks), intermediate (24 weeks), and long-term (52 weeks).
📈 Results
The study identified several key clinical and laboratory parameters that significantly impacted overall survival. Notably, the AUC values for the predictive models were impressive: 0.901 for short-term OS, 0.767 for intermediate OS, and 0.806 for long-term OS. The inclusion of IL-8 in both short- and long-term models highlights its importance as a prognostic factor.
🌍 Impact and Implications
The development of these predictive models represents a significant advancement in the management of patients with bone metastases. By utilizing readily available clinical and laboratory data, healthcare providers can make more informed decisions regarding palliative RT, ultimately improving patient outcomes. However, further studies are essential to validate these findings and facilitate their integration into clinical practice.
🔮 Conclusion
This study underscores the potential of predictive modeling in enhancing palliative care for patients with bone metastases. The promising results of the developed models pave the way for future research and clinical applications, aiming to provide better prognostic tools for healthcare professionals. Continued exploration in this area is vital for improving patient care and outcomes.
💬 Your comments
What are your thoughts on the use of predictive models in palliative care? We would love to hear your insights! 💬 Share your comments below or connect with us on social media:
Development of a predictive model for patients with bone metastases referred to palliative radiotherapy: Secondary analysis of a multicenter study (the PRAIS trial).
Abstract
BACKGROUND: The decision to administer palliative radiotherapy (RT) to patients with bone metastases (BMs), as well as the selection of treatment protocols (dose, fractionation), requires an accurate assessment of survival expectancy. In this study, we aimed to develop three predictive models (PMs) to estimate short-, intermediate-, and long-term overall survival (OS) for patients in this clinical setting.
MATERIALS AND METHODS: This study constitutes a sub-analysis of the PRAIS trial, a longitudinal observational study collecting data from patients referred to participating centers to receive palliative RT for cancer-induced bone pain. Our analysis encompassed 567 patients from the PRAIS trial database. The primary objectives were to ascertain the correlation between clinical and laboratory parameters with the OS rates at three distinct time points (short: 3 weeks; intermediate: 24 weeks; prolonged: 52 weeks) and to construct PMs for prognosis. We employed machine learning techniques, comprising the following steps: (i) identification of reliable prognostic variables and training; (ii) validation and testing of the model using the selected variables. The selection of variables was accomplished using the LASSO method (Least Absolute Shrinkage and Selection Operator). The model performance was assessed using receiver operator characteristic curves (ROC) and the area under the curve (AUC).
RESULTS: Our analysis demonstrated a significant impact of clinical parameters (primary tumor site, presence of non-bone metastases, steroids and opioid intake, food intake, and body mass index) and laboratory parameters (interleukin 8 [IL-8], chloride levels, C-reactive protein, white blood cell count, and lymphocyte count) on OS. Notably, different factors were associated with the different times for OS with only IL-8 included both in the PMs for short- and long-term OS. The AUC values for ROC curves for 3-week, 24-week, and 52-week OS were 0.901, 0.767, and 0.806, respectively.
CONCLUSIONS: We successfully developed three PMs for OS based on easily accessible clinical and laboratory parameters for patients referred to palliative RT for painful BMs. While our findings are promising, it is important to recognize that this was an exploratory trial. The implementation of these tools into clinical practice warrants further investigation and confirmation through subsequent studies with separate databases.
Author: [‘Rossi R’, ‘Medici F’, ‘Habberstad R’, ‘Klepstad P’, ‘Cilla S’, “Dall’Agata M”, ‘Kaasa S’, ‘Caraceni AT’, ‘Morganti AG’, ‘Maltoni M’]
Journal: Cancer Med
Citation: Rossi R, et al. Development of a predictive model for patients with bone metastases referred to palliative radiotherapy: Secondary analysis of a multicenter study (the PRAIS trial). Development of a predictive model for patients with bone metastases referred to palliative radiotherapy: Secondary analysis of a multicenter study (the PRAIS trial). 2024; 13:e70050. doi: 10.1002/cam4.70050