Table 4.

Correlation between relapse/residual disease (prognosis) and/or death (survival) with QSOX1 expression, tumor differentiation, tumor stroma pattern and Shimada classification. Sensitivity, specificity, positive predictive value and negative predictive value were determined.

	Relapse/residual disease and/or death		Total	P value	Quality score	Results (confidence intervals)
	Yes	No
QSOX >65	10 (58.8%)	0 (0%)	10		Sens	58.8 (35.4-82.2)
QSOX ≤65	7 (41.2%)	6 (100%)	13	0.019	Spec	100 (100-100)
					PPV	100 (100-100)
Total	17	6	23		NPV	46.2 (19.1-73.3)
Poor-differentiated tumor	5 (29.4%)	4 (66.7%)	9		Sens	29.4 (7.8-51.1)
Well-differentiated tumor	12 (70.6%)	2 (33.3%)	14	0.162	Spec	33.3 (0-71.1)
					PPV	55.6 (23.1-88.0)
Total	17	6	23		NPV	14.3 (0-32.6)
Stroma-poor tumor	5 (29.4%)	4 (66.7%)	9		Sens	29.4 (7.8-51.1)
Stroma-rich tumor	12 (70.6%)	2 (33.3%)	14	0.162	Spec	33.3 (0-71.1)
					PPV	55.6 (23.1-88.0)
Total	17	6	23		NPV	14.3 (0-32.6)
Shimada unfavourable	4 (23.5%)	1 (16.7%)	5		Sens	23.5 (3.4-43.7)
Shimada favourable	13 (76.5%)	5 (83.3%)	18	1	Spec	83.3 (53.5-100)
					PPV	80 (44.9-100)
Total	17	6	23		NPV	27.8 (7.1-48.5)

Sens, sensitivity: P (QSOX>65 µm² as occurs in relapse/residual disease and/or death); Spec, specificity: P (QSOX≤65 µm² because there is relapse/residual disease and/or death); PPV, positive predictive value: P (a relapse/residual disease and/or death as QSOX>65 µm²); NPV, negative predictive value: P (not a relapse/residual disease and/or death as QSOX≤65 µm²). Mean immunopositivity for QSOX1 for each sample was measured in µm² per high-power field. °Statistically significant value.