Skip to main content
. 2021 Mar 8;21:250. doi: 10.1186/s12885-021-07948-9

Table 3.

Predictors for overall survival after primary curative treatment by using Cox regression model

Variables Univariate model aMultivariate model
HR (95% CI) Model I
Time-invariant without recurrence
aHR (95% CI)
Model II
Time-invariant with recurrence
aHR (95% CI)
Model III
Time-varying with recurrence
aHR (95% CI)
Time-dependent variables
 Recurrence type
  No 1.00 1.00
  Early recurrence (< 1 year) 9.44 (4.98–17.9) 6.62 (3.79–11.6)
  Late recurrence (> 1 year) 6.46 (3.17–13.2) 3.75 (1.99–7.08)
Time-independent variables
 Recurrence type
  No 1.00 1.00
  Early (< 1 year) 4.51 (2.42–8.39) 3.09 (2.03–4.71)
  Late (> 1 year) 1.73 (0.91–3.28)
 Host-related factors
  Age > 65 years 1.64 (1.09–2.48) 1.43 (0.92–2.21) 1.55 (1.01–2.36)
  Male 1.23 (0.81–1.86)
  AST > 80 IU/L 1.30 (0.80–2.11)
  ALT > 80 IU/L 0.99 (0.60–1.63)
  Total bilirubin > 1.5 mg/dL 1.93 (1.00–3.72)
  Prothrombin time (INR) > 1.3 1.86 (0.86–4.04)
  Albumin <3.0 mg/dL 2.38 (1.20–4.75)
  Child-Pugh score B (vs A) 1.63 (0.75–3.53)
  CSPH 2.31 (1.50–3.56) 1.53 (0.93–2.52) 1.71 (1.04–2.82) 1.97 (1.26–3.08)
  Creatinine > 1.5 mg/dL 2.04 (0.98–4.22) 1.93 (0.92–4.02) 1.89 (0.90–3.97)
  Diabetes 1.23 (0.79–1.90)
  ALBI grade 2 or 3 (vs 1) 1.91 (1.02–3.59)
  FIB-4 index > 3.25 2.50 (1.63–3.84) 2.02 (1.23–3.31) 1.83 (1.12–3.01)
 Tumor characteristics
  Size > 3 cm 1.02 (0.67–1.54)
  Multinodularity 1.18 (0.75–1.84)
  AFP > 20 ng/mL 1.20 (0.79–1.81)
  BCLC stage A (vs 0) 1.94 (1.03–3.65) 1.88 (1.00–3.55)
 Histopathological findings
  Edmonson grade III & IV 1.35 (0.76–2.40)
  Ishak fibrosis score 4–6 1.24 (0.79–1.96)
 Viral factors
  HBV 5.23 (2.86–9.57)
  HCV 5.36 (2.96–9.70)
  HBV and HCV 10.3 (3.77–28.1)
 Treatment modality
  RFA-TACE (vs. Surgery) 1.25 (0.82–1.91)
Model selection
 -2 log likelihood ratio 871.9 847.7 834.5
 AIC 879.9 857.7 842.5

aHR adjusted hazard ratio, ALT alanine aminotransferase, AST aspartate aminotransferase, INR international normalized ratio, CSPH clinically significant portal hypertension, ALBI albumin-bilirubin, AFP alpha-fetoprotein, HBV hepatitis B virus, HCV hepatitis C virus, BCLC Barcelona-Clınic Liver Cancer staging system, RFA radiofrequency ablation, AIC Akaike information criterion

aMultivariate analyses were performed by the Cox proportional model with forward selection, with P < 0.15 indicating inclusion and removal for variable selection. Model I: variables without recurrence. Model II: recurrence as time-invariant variables. Model III: recurrence as time-varying variables