Table 2. Descriptive statistic, univariate and multivariate analyses for PRL-3 and clinicopathological variables.

	Descriptive			Univariate Cox's proportional hazard model			Univariate binary logistic regression			Multivariate Cox's proportional hazard model			Multivariate binary logistic regression
	N	%	Mean survival	HR	CI	P-value	OR	CI	P-value	HR	CI	P-value	OR	CI	P-value
Gender
Female	40	50	78.76±6.8	1	0.672–2.948	0.366	1	0.683–4.022	0.263	1	0.54–2.46	0.714	1	0.483–4.072	0.533
Male	40	50	67.04±5.8	1.407			1.658			1.153			1.403
Tumour stage
IIIB	46	57.5	80.11±5.53	1	0.951–4.121	0.068	1	0.983–6.044	0.055	1	0.848–4.076	0.122	1	0.793–7.9	0.118
IIIC	34	42.5	63.14±7.26	1.98			2.437			1.859			2.503
Location
Colon	54	67.5	76.09±7.38	1	0.544–2.625	0.657	1	0.393–2.51	0.99	1	0.564–2.794	0.578	1	0.327–3.118	0.987
Rectum	26	32.5	71.36±5.68	1.195			1.006			1.255			1.009
Vascular or lymphatic invasion
No	39	48.8	82.63±6.01	1	0.866–3.886	0.113	1	0.928–5.607	0.072	1	0.484–2.544	0.805	1	0.392–3.792	0.732
Yes	41	51.3	63.64±6.41	1.834			2.282			1.110			1.219
PRL-3 expression
Low	42	52.5	87.12±5.75	1	1.607–8.24	0.002	1	3.65–28.81	<0.0001	1	1.405–7.852	0.006	1	3.308–28.98	<0.0001
High	38	47.5	57.17±6.16	3.639			10.267			3.322			9.791

Cox's proportional hazard models were fitted to estimate hazard ratios (HRs), and 95% confidence interval (CI) and likelihood ratio tests were performed to assess the statistical significance of the variables.

Binary logistic regression was performed to identify predictors for distant dissemination.