Natural disease progression and novel survival prediction model for hepatocellular carcinoma with spinal metastases: a 10-year single-center study

Abstract

Background: Individual prediction of life expectancy in patients with spinal metastases from hepatocellular carcinoma (HCC) is key for optimal treatment selection, especially when identifying potential candidates for surgery. Most reported prognostic tools provide categorical predictions, and only a few include HCC-related factors. This study aimed to investigate the natural progression of the disease and develop a prognostic tool that is capable of providing individualized predictions.

Methods: Patients with HCC-derived metastatic spinal disease were identified from a retrospective cohort of patients with spinal metastases who were diagnosed at Chiang Mai University Hospital between 2006 and 2015. Kaplain-Meier methods and log-rank tests were used to statistically evaluate potential factors. Significant predictors from the univariable analysis were included in the flexible parametric survival regression for the development of a prognostic prediction model.

Results: Of the 1143 patients diagnosed with HCC, 69 (6%) had spinal metastases. The median survival time of patients with HCC after spinal metastases was 79 days. In the multivariable analysis, a total of 11 potential clinical predictors were included. After backward elimination, four final predictors remained: patients aged > 60 years, Karnofsky Performance Status, total bilirubin level, and multifocality of HCC. The model showed an acceptable discrimination at C-statistics 0.73 (95% confidence interval 0.68-0.79) and fair calibration.

Conclusion: Four clinical parameters were used in the development of the individual survival prediction model for patients with HCC-derived spinal metastases of Chiang Mai University or HCC-SM CMU model. Prospective external validation studies in a larger population are required prior to the clinical implication of the model.

Keywords: Decision making; Hepatocellular carcinoma; Prognosis; Prognostic factors; Spinal metastases; Survival analysis.

Fig. 1

Kaplan–Meier curves visualizing differences in survival distribution among patients with and without prognostic factors

Fig. 2

Calibration plots compare the model-predicted risk and the observed outcomes against one another within each of the risk quantiles

Fig. 3

Individual prediction curves of nine randomly selected patients from the study cohort

Fig. 4

HCC-SM CMU score chart. Each patient’s predicted survival probabilities at each clinically relevant time point (3, 6, and 12 months) classified into five risk groups with a specific coloring label as follows: 81–100% (green), 61–80% (yellow), 41–60% (light orange), 21–40% (dark orange), and 0–20% (pink). Abbreviation: KPS, Karnofsky Performance Status