Predictive Scores Underestimate Survival of Patients With Metastatic Spine Disease

A Retrospective Study of 315 Patients in Sweden

Christian Carrwik, MD; Claes Olerud, MD, PhD; Yohan Robinson, MD, PhD, MBA


Spine. 2020;45(6):414-419. 

In This Article



The search in Swespine yielded 316 patients, of which one was excluded from further analysis due to emigration and thus lacking reliable survival data. For the 315 included patients (213 men, 102 women, mean age 62 years), we gathered further data from their medical records at Uppsala University Hospital. Prostate cancer accounted for 27% of the metastasis, whereas primary tumor was unknown before surgery in 22% of the cases (Table 1).

The most common indication for surgery in this series was neurologic deficit (80% of the cases), whereas pain was the second most common (17%). Progressing deformity (1%) was the least common indication, whereas data on indication are missing in 2% (n = 7) of the cases. Posterior decompression and stabilization was the most common surgical method, followed by anterior decompression and stabilization (Table 2). Th3 was the most common proximal level of decompression, whereas Th8 was the most common rostral level of decompression.

Outcome Data

The mean estimated survival time after surgery for all patients included was 12.4 months (confidence interval 10.6–14.2) and median survival 5.9 months (confidence interval 4.5–7.3). All four scores has significant correlation to survival (P < 0.0001).

Twenty-eight percent of the patients were dead within 3 months after surgery, whereas 55% were alive 6 months after surgery. Approximately 39% were alive more than 12 months after surgery.

Main Results

All of the scoring systems predicted a lower rate of long-surviving patients, but even overestimating of survival was seen. According to the original Tokuhashi score, 99 of the patients were scored in the group with the highest expected survival. In contrast, only 18 were scored in the best group according to the Tomita score. Of those patients with the longest expected survival according to Tomita Score 4 out of 18 died within 3 months after surgery, which equals to 22% overestimation with our definition. The Revised Tokuhashi score overestimated only 7% of the patients with long estimated survival (Table 3).

Modified Bauer score was the best of the four scores to predict short survival, both regarding mean and median survival. Patients who had expected long survival according to Revised Tokuhashi score had the longest median survival. However, the Revised Tokuhashi score had the highest proportion of underestimated survival among the investigated scores and could be regarded as the most cautious score in this material (Table 4).

All scoring systems underestimated rather than overestimated survival with our definitions of accuracy in the estimations (Figure 1).

Figure 1.

Diagram showing the distribution between overestimated and underestimated survival among different scoring systems. All scores tend to underestimate rather than overestimate survival, according to our definitions.