Preventing Discontinuation of Radiation Therapy: Predictive Factors to Improve Patient Selection for Palliative Treatment.

PURPOSE: Approximately one third of patients with cancer require palliative radiation therapy (PRT), yet no guidelines exist for optimal patient selection. We have observed that many patients who begin PRT do not complete their prescribed treatment. Our study sought to identify factors associated with discontinuation of PRT, assess for a relationship with survival, and inform patient selection. METHODS: We performed an institutional review board-approved retrospective analysis of patients with cancer treated in a multicenter radiation oncology department in 2014. Of 297 patients who began PRT, 60 discontinued and 237 completed treatment. Primary end points included discontinuation and overall survival. RESULTS: Patient factors were analyzed for association with discontinuation of PRT and overall survival, respectively, using logistic regression and Cox proportional regression models. Factors associated with discontinuation were low Karnofsky performance status (KPS) score, high number of fractions prescribed, and treatment site other than bone metastasis. The odds of discontinuing PRT decreased by approximately 52% for every 10-point increase in KPS score (odds ratio, 0.48; 95% CI, 0.36 to 0.63; P < .001). Factors associated with shorter survival included discontinuation of PRT, low KPS score, community practice location, multiple comorbidities, and treatment of brain metastases. Patients who discontinued treatment were more likely to die than patients who completed treatment, independent of other factors (hazard ratio, 3.67; 95% CI, 2.41 to 5.61; P < .001). CONCLUSION: Patients with low KPS scores, long treatment courses, and those treated to sites other than bone metastasis were significantly more likely to discontinue treatment. Discontinuation was predictive for poor survival. Pretreatment evaluation of KPS, comorbidities, and brain metastases can help guide appropriate patient selection for PRT.