Table 2.

Multivariate analysis for disease-specific survival

	Wald χ2	HR	95% CI	p value
Training cohort
mvi	48.2	2.41	1.88–3.09	<0.0001
Cirrhosisa	30.8	2.01	1.57–2.57	<0.0001
Multiple lesions	11.2	1.69	1.24–2.31	0.0008
Size >2 cm	7.1	1.38	1.09–1.76	0.0078
Poor differentiationc	6.9	1.44	1.10–1.90	0.0085
Age >65	5.4	1.32	1.05–1.68	0.0201
Validation cohort
mvi	17.6	1.65	1.30–2.08	<0.0001
Multiple lesions	15.4	1.67	1.29–2.15	<0.0001
Poor differentiationc	14.3	1.59	1.25–2.02	0.0002
Cirrhosisb	12.6	1.47	1.19–1.82	0.0004
Size >2 cm	11.7	2.07	1.36–3.14	0.0006

Multivariate Cox-hazard regression was applied with stepwise backward selection. Initially, all factors were included in the model. Then factors that showed no or limited statistically significant association (p > 0.1) with disease-specific survival adjusted for the remaining factors were deleted from the model in stepwise fashion. The 7 factors tested were as follows: age (> vs. ≤65), gender, maximum tumor size (> vs. ≤2 cm), number of tumor (multiple vs. solitary), microvascular invasion (presence vs. absence), cirrhosis (presence vs. absence), and tumor differentiation (poor vs. well/moderate). HR, hazard ratio; mvi, microvascular invasion.

^aDesmet's classification [13] grade 4.

^bIshak score [12] F5 or F6.

^cEdmondson grade 3 or 4 [11].