TL;DR
This study evaluates prognostic factors for patients with five or more brain metastases treated with stereotactic radiosurgery (SRS) instead of whole-brain radiation. The researchers found that while both the number and volume of new metastases over time correlate with survival, the simple count of new lesions (Brain Metastasis Velocity) is a more robust predictor of patient outcomes than the total volume of those lesions.
What Problem Does This Paper Address?
When cancer spreads to the brain (metastasis), clinicians traditionally used Whole-Brain Radiotherapy (WBRT). However, WBRT is associated with significant neurocognitive decline and reduced quality of life. Stereotactic Radiosurgery (SRS)—a highly targeted, high-dose radiation technique—has become the standard for patients with 1–4 metastases, but its use for patients with five or more lesions is still debated.
The challenge lies in patient selection: which patients with extensive brain involvement will actually benefit from SRS, and who might be better served by supportive care or WBRT? The authors sought to identify clinical markers—specifically comparing the "velocity" of new metastases by count (BMV) versus volume (vBMV)—to help guide these difficult treatment decisions.
Methods
The authors conducted a retrospective analysis of 116 patients treated at the University of Rochester between 2019 and 2024. All patients had $\ge 5$ brain metastases (median of 10, ranging up to 41) and received linear accelerator-based SRS without prior whole-brain radiation.
Key methodological features included:
- Prognostic Scoring: Use of Graded Prognostic Assessments (GPA), a standardized tool that predicts survival based on age, performance status, and cancer type.
- Velocity Metrics: The study calculated two types of "velocity" for patients who developed new lesions after their first SRS:
- BMV: Number of new metastases per year.
- vBMV: Total volume (in cc) of new metastases per year.
- Statistical Modeling: Univariable and multivariable Cox regression models were used to determine which factors independently predicted Overall Survival (OS).
Datasets
The study utilized a single-institution clinical dataset from the University of Rochester.
- Cohort size: 116 patients.
- Primary cancers: Non-small cell lung cancer (54.3%), melanoma (17.2%), breast (16.4%), and others.
- Follow-up: Median follow-up was 8.2 months, with some patients followed for over 6 years.
Key Insights and Ideas
- The "Velocity" Showdown: While one might assume the volume of new disease (vBMV) would be more dangerous than the number of spots (BMV), this study found that BMV was a more significant prognostic factor in multivariable models. This suggests that the frequency of new "seeding" events in the brain is a stronger indicator of aggressive biology than the sheer size of the new tumors.
- Identifying "Poor Survivors": About 18% of the cohort died within two months of treatment. These patients were more likely to have progressive disease elsewhere in the body (extracranial) and lower GPA scores. This insight helps clinicians identify patients where SRS might be "over-treatment" compared to supportive care.
- Extracranial Control is King: The status of the cancer outside the brain at the time of SRS was one of the strongest predictors of survival. If the primary cancer is progressing rapidly, treating the brain metastases with SRS may not significantly extend life.
Key Results
- Overall Survival: The median OS was 8.2 months. Survival rates were 60.3% at 6 months, 40.5% at 12 months, and 28.0% at 24 months.
- BMV Significance: For patients who survived long enough for follow-up, BMV was significantly associated with survival ($p = 0.018$ on multivariable analysis). Each additional lesion per year increased the hazard ratio for death by approximately 1.7%.
- GPA Validation: GPA-predicted survival correlated well with actual outcomes ($p < 0.0001$), confirming its utility even in patients with high lesion counts.
- Local Control: SRS was highly effective at the treated sites; only 13.5% of patients experienced a local recurrence of an original lesion.
Limitations and Future Work
The study is limited by its retrospective nature and single-institution scope, which can introduce selection bias (e.g., the doctors may have only offered SRS to patients they already thought would do well). Additionally, the sample sizes for specific cancer types like melanoma or kidney cancer were too small to draw definitive, histology-specific conclusions.
The authors suggest that future work should focus on multi-institutional cohorts to refine these selection criteria. They also highlight the potential for incorporating radiomic and genomic data alongside artificial intelligence to create more precise predictive models for brain metastasis behavior.
Summary updated