Development and Internal Validation of a Recursive Partitioning Analysis-Based Model Predictive of Pain Flare...

Pain flares are a common acute toxic effect after stereotactic body radiation therapy (SBRT) for spine metastasis. We aimed to identify a subset of patients with the highest rate of pain flare after spine SBRT to optimize prophylactic corticosteroid administration.

The data set included 428 patients with 610 treatments. We defined pain flare as acute worsening of pain at the treatment site requiring new or higher dose therapy with corticosteroids, opiates, and/or hospitalization. Data were split into 70% training and 30% validation sets using a random number generator. After feature importance testing and generation of a correlation heatmap, feature extraction was performed via recursive partitioning analysis.

We identified 125 total pain flares (20%). Five variables met significance (P < .02) for model inclusion: renal primary, soft tissue involvement, Bilsky >0, spinal instability neoplastic score >6, and gross tumor volume >8 cc. One point was assigned for each variable. The low-risk group (score = 0, n = 159) had pain flare rates of 7.0% and 13.6% in the training and validation sets; the intermediate-risk group (score = 1, n = 150) had rates of 14.0% and 16.3%; and the high-risk group (score >1, n = 301) had rates of 28.8% and 31.3%. Patients in the high-risk group had higher rates of flare (odds ratio, 3.50; 95% confidence interval, 2.06-5.92) and accumulated health care costs 3 and 6 months post-SBRT, relative to intermediate- and low-risk patients (P < .001).

Our internally validated model identifies a high-risk group of patients more likely to develop a pain flare after spine SBRT, for whom prophylactic steroids may be considered. Evaluation in a clinical trial is warranted.