A Prognostic Model for Patients With Oligometastatic Disease Treated With Stereotactic Body Radiation Therapy

Abstract

Listen

Translate article

<h3>Purpose/Objective(s)</h3> Stereotactic body radiation therapy (SBRT) is an increasingly important modality in the management of patients with oligometastatic disease (OMD). Though prospective clinical trials have demonstrated the benefits of SBRT in a variety of oligometastatic settings, there is currently limited data to guide patient selection and provide long-term prognostic information for OMD patients. The purpose of this study was to create a clinical prognostic model for overall survival (OS) for OMD patients treated with SBRT. <h3>Materials/Methods</h3> A large, retrospective multi-institutional database of OMD patients treated with SBRT provided the data for model construction. Recursive partitioning analysis (RPA) was used to generate a prognostic model for OS that could account for complex interactions between baseline patient characteristics. The model was generated using a training set (75% of all samples) and internally validated using the reserved testing set. Model performance in the training and test sets were evaluated using log-rank tests, Harrell's C-statistic and time-dependent area under the receiver operating characteristics curve (AUC). All analyses were carried out in R. <h3>Results</h3> A total of 1,033 patients were included in the analysis. RPA for OS revealed three risk groups. The low-risk group consisted of younger (< 55) patients with favorable primary sites (hormone receptor/Her2-positive breast cancer, colorectal cancer or renal cell carcinoma) as well as any patient with a prostate cancer primary; the high-risk group consisted of patients with any other primary site who presented with non-pulmonary OMD within 24 months of the diagnosis of the primary disease; and the intermediate-risk group consisted of all other patients. The 5-year OS was 77.5 % (95% confidence interval: 63.1-91.9%), 32.3% (25.1-39.5%) and 12.1% (2.9-21.4%), respectively, for the low, intermediate and high-risk groups. Log-rank tests for difference in survival between the risk groups in both the training and test sets were highly significant (<i>P</i> < 0.0001). The model possessed good discriminative power with a C-statistic of 0.68 and time-dependent AUC of 0.72 in the training set, and there was an expected small reduction in these statistics in the test set (C-statistic: 0.65, AUC: 0.67). <h3>Conclusion</h3> An internally validated prognostic model for OS with good ability to distinguish between low, intermediate and high-risk OMD patients was generated. Subsequent external validation will be undertaken to demonstrate the robustness of this model to new data.