A Survival Prediction Tool for Cancer Patients Treated with Stereotactic Radiation Therapy

Purpose/Objective(s)

We propose a predictive model that identifies patients at greatest risk of death after receiving stereotactic radiotherapy as part of their cancer treatment.

Materials/Methods

The National Cancer Database was utilized to identify patients who received stereotactic radiotherapy during their first course of cancer treatment, regardless of TNM stage. Dose ranged from 16-26 Gy in 1 fraction to 30-60 Gy in 5 fractions. Survival predictors, mortality risk groups and risk indices were assessed using Cox regression models. Adjusted hazard ratios (HRs) with 95% confidence intervals (CI) and p-values based on Wald tests were reported for: 1) age 2) gender 3) Charlson-Deyo comorbidity score 4) metastases location (liver, bone, lung, brain) 5) tumor primary (prostate, breast, lung, other) and 6) radiotherapy site.

Results

A total of 56,292 patients were included from 2010-2014. Median age was 71.0 years, 22% had metastases, and 71% of patients were treated for localized lung cancer. Adjusted HRs for the survival predictors are in Table 1. The median survival times were 79.8 months, 33.1 months, and 15.5 months in the low- (n=18,564), intermediate- (n=18,647), and high-risk groups (n=18,489), respectively. Higher survival was associated with primary prostate or breast cancer (compared to lung or other primaries) and pelvis and breast radiation sites. Patients receiving stereotactic radiation to the brain and eye had the lowest median survival. A nomogram was created to predict survival. Harrel's C-index was 0.6676 and area under the curve was 0.6589 at 4 years.

Conclusion

We created a predictive survival nomogram for cancer patients receiving stereotactic radiotherapy based on age, sex, morbidity, and site of metastases, radiation, and primary tumor. The model was implemented as an online tool (.https://tinyurl.com/rad-surv) that can provide improved decision-making process and treatment stratification.