Modeling 1-year survival after surgery on the metastatic spine

Background context: Choosing appropriate surgical patients in the setting of spinal metastases can be challenging. Existing scoring systems focus primarily on patient selection or operative techniques. These scores are limited in their capacity to predict postoperative survival.

Purpose: The aim was to model survival after spine surgery for metastastic disease.

Study design: This was a retrospective multicenter study.

Patient sample: All patients who had undergone surgery for the treatment of metastatic spinal disease at one of four tertiary care centers between 2007 and 2013 were included.

Outcome measure: The outcome measure was 1-year survival after surgery.

Methods: Demographic, medical, oncologic, surgical, and survival data were abstracted from medical records. The effect of predictor variables on survival was evaluated alone and in combination using stepwise logistic regression. Multivariable logistic regression was subsequently used to adjust for confounders. A predictive score was then developed and compared against that of the modified Bauer score alone in terms of prognosticating 1-year survival after surgery.

Results: In the time period under investigation, 318 patients underwent surgical intervention for metastastic disease involving the spine, with 307 having data available for analysis. The survival rate at 1 year was 48% (n=142), with a median survival of 10 months. In final adjusted analysis, preoperative modified Bauer score (odds ratio [OR] 3.00; 95% confidence interval [CI] 1.80-5.01; p<.001), ambulatory status (OR 2.47; 95% CI 1.48-4.14; p=.001), and serum albumin (OR 2.80; 95% CI 1.66-4.72; p<.001) were all independent predictors of 1-year survival. The most parsimonious model weighted the modified Bauer score with 2 points and intact ambulatory status and normal serum albumin level with 1 point each, with a ceiling score of 3. The final model using the predictive score was able to explain 74% of the variation in 1-year survival. In contrast, the modified Bauer score alone was only able to explain 64% of the variation in 1-year survival.

Conclusions: This study demonstrates the importance of including factors related to the overall health of a patient, in addition to parameters surrounding their cancer diagnosis, to better prognosticate survival. Our predictive score performed better than the modified Bauer alone and may be used to predict survival after surgical intervention for metastatic disease.

Level of evidence: III.