Survival Prediction in Patients Treated Surgically for Metastases of the Appendicular Skeleton-An External Validation of 2013-SPRING Model.

Abstract

Simple SummaryBone tumour metastases are highly prevalent among cancer patients. In case these have to be treated surgically due to impending or pathological fracture, estimation of patients’ life expectancy is of importance in order to choose the best treatment option possible. In the current study, the 2013-SPRING model, developed to predict prognosis of surgically treated bone metastasis patients, was validated in an external patient cohort of 303 bone metastasis patients. AUC ROCs at all three endpoints assessed (i.e., survival at 3, 6 and 12 months following surgery for bone metastases) were all above 0.780. Furthermore, stratification into type of surgery (endoprosthesis (n = 162) vs. osteosynthesis (n = 141) and metastasis location (upper limb (n = 65) vs. lower limb (n = 238)) revealed a comparable predictive accuracy of the 2013-SPRING model, albeit slightly better performance in the osteosynthesis as compared with endoprosthesis subgroup, as well as upper limb in comparison to lower limb subgroup was observed.Introduction: The aim of this study was to externally validate the 2013-SPRING model, a survival prediction tool for patients treated surgically for bone metastases in a retrospective patient cohort from a single institution. Moreover, subgroup analyses on patients treated with (A) endoprostheses or (B) osteosynthesis, as well as (C) upper limb and (D) lower limb metastases, were performed. Methods: Altogether, 303 cancer patients (mean age: 67.6 ± 11.1 years; 140 males (46.2%)) with bone metastases to the extremities, treated surgically between March 2000 and June 2018 at a single tertiary sarcoma centre, were retrospectively included. Median follow-up amounted to 6.3 (interquartile range (IQR): 2.3–21.8) months, with all patients followed-up for at least one year or until death. The 2013-SPRING model was applied to assess the prognostication accuracy at 3, 6 and 12 months. Models were validated with area under the curve receiver operator characteristic (AUC ROC; the higher the better), as well as Brier score. Results: Of the 303 patients, 141 had been treated with osteosynthesis (46.5%), and the remaining 162 patients with endoprosthesis (53.5%). Sixty-five (21.5%) metastases were located in the upper limbs, and two hundred and thirty-eight (78.5%) in the lower limbs. Using the 2013-SPRING model for the entire cohort, the accuracy of risk of death prediction at 3, 6 and 12 months, determined by the AUC ROC, was 0.782 (95% CI: 0.729–0.843), 0.810 (95% CI: 0.763–0.858) and 0.802 (95% CI: 0.751–0.854), respectively. Corresponding Brier scores were 0.170, 0.178 and 0.169 at 3, 6 and 12 months. In the subgroup analyses, predictive accuracy of the 2013-SPRING model was likewise encouraging, albeit being slightly higher in the osteosynthesis subgroup as compared with the endoprosthesis subgroup, and also higher in the upper limb in comparison to the lower limb metastasis subgroup. Conclusions: The current validation study of the 2013-SPRING model shows that this model is clinically relevant to use in an external cohort, also after stratification for surgical procedure and metastasis location.