Validation of nine prognostic models in patients with bone metastases undergoing surgery

Background: We aimed to 1) assess the prognostic performance of established prediction models for mortality in patients with spine and extremity bone metastases, 2) evaluate their performance in clinically relevant subgroups, and 3) study the importance of relevant patient characteristics for mortality prediction. Methods: Between 29th May, 2000 and 21st April, 2022, a total of 526 patients (median age 67.0 years, interquartile range [IQR] 58.1 - 74.0 years], 271 males; median follow-up 267 days [IQR 86 - 856 days]) undergoing surgery for spine or extremity bone metastases were retrospectively included. Nine prognostic models were evaluated in the entire cohort (n = 526), and separately in subgroups treated for spine (n = 224) or extremity (n = 302) metastases. Cox-regression and logistic regression models were used, as appropriate. Harrell's c-statistic, the area under the receiver operating characteristic curve (AUC ROC) and Brier score were used as performance metrics. Results: When assessed in the entire cohort, models by Sorensen et al. (AUC 12 months 0.834), Janssen et al. (c-index 0.711) and Katagiri et al. (c-index 0.699) achieved highest discriminatory performance. Likewise, all three models performed best when studied in the spine subgroup (Sorensen et al.: AUC 12 months 0.826; Janssen et al.: c-index 0.723; Katagiri et al.: c-index 0.717), although the models by Sorensen et al. and Janssen et al. had been developed for patients with extremity metastasis. Performance of all models was slightly lower when assessed in the extremity subgroup. Haemoglobin levels and tumour profile (i.e. primary histology) ranked among the most important predictors, consistent across subgroups. Conclusions: Our study suggests particularly helpful predictive performance of the models by Sorensen et al. and Janssen et al. in both patients with extremity metastases and patients with spine metastases undergoing surgery. Given that we only included patients undergoing surgery, additional validation studies in conservatively treated patients are warranted.