Predicting malignant transformation of esophageal squamous cell lesions by combined biomarkers in an endoscopic screening program

Table 1 Distribution of selected variables in the esophageal basal cell hyperplasia, low-grade dysplasia, high-grade dysplasia, esophageal squamous cell cancer and control groups¹n (%)

Variables	ESCC	HGD	LGD	BCH	Normal control
Gender
Male	89 (66.4)	32 (62.7)	146 (59.1)	48 (62.3)	40 (50.0)
Female	45 (3.6)	19 (37.3)	101 (40.9)	29 (37.7)	40 (50.0)
Age (yr)
40-50	52 (38.8)	16 (31.4)	60 (24.3)	20 (26.0)	7 (8.8)
50-60	60 (44.8)	25 (49.0)	128 (51.8)	34 (44.2)	29 (36.3)
≥ 60	22 (16.4)	10 (19.6)	59 (23.9)	23 (29.9)	44 (55.0)
School year (yr)
≤ 6	79 (59.0)	20 (39.3)	105 (42.5)	32 (41.6)	23 (28.8)
7-11	50 (37.3)	22 (43.1)	98 (39.7)	30 (39.0)	36 (45.0)
≥ 12	5 (3.7)	9 (17.6)	44 (17.8)	15 (19.5)	21 (26.3)
Income per year-person ($)
< 150	60 (44.8)	22 (43.1)	101 (40.9)	26 (33.8)	8 (10.0)
150-350	56 (41.8)	17 (33.3)	92 (37.2)	24 (31.2)	21 (26.3)
≥ 350	18 (13.4)	12 (23.5)	54 (21.9)	27 (35.1)	51 (63.8)
Family history of esophageal cancer
Yes	19 (14.2)	12 (23.5)	50 (20.2)	16 (20.8)	8 (10.0)
No	115 (85.8)	39 (76.5)	197 (79.8)	61 (79.2)	72 (90.0)
Smoking index2
≥ 450	52 (38.8)	23 (45.1)	69 (27.8)	31 (40.0)	14 (17.7)
< 450	25 (18.7)	4 (7.8)	40 (16.3)	12 (16.0)	16 (20.3)
None	57 (42.5)	24 (47.1)	138 (55.9)	34 (44.0)	49 (62.0)
Alcohol drinking index3
≥ 120	56 (42.0)	16 (31.4)	77 (31.0)	22 (28.0)	6 (7.5)
< 120	28 (20.6)	11 (21.6)	44 (18.0)	17 (22.7)	29 (36.3)
None	50 (37.4)	24 (47.1)	126 (51.0)	38 (49.3)	45 (56.3)

¹In the five groups, χ² test values for gender, age, school year, family history of esophageal cancer, income per year-person, smoking index, and alcohol drinking index were 6.046 (P = 0.196), 52.858 (P < 0.01), 45.436 (P < 0.01), 76.476 (P < 0.01), 7.137 (P = 0.129), 21.682 (P = 0.006), and 34.085 (P < 0.01), respectively;

²Smoking index = cigarette/day × number of smoking years;

³Alcohol drinking ≥ 120 g/d represents heavy drinking. ESCC: Esophageal squamous cell cancer; HGD: High-grade dysplasia; LGD: Low-grade dysplasia; BCH: Basal cell hyperplasia.

Table 2 Distribution of p53, CA19-9 and carcinoembryonic antigen expression in the five groups¹n (%)

Protein expression	ESCC	HGD	LGD	BCH	Normal control
p53
Negative	63 (47.0)	24 (47.1)	159 (64.4)	56 (72.7)	64 (80.0)
Positive
1%-10%	18 (13.4)	17 (33.3)	62 (25.1)	17 (22.1)	11 (13.8)
10%-25%	20 (14.9)	8 (15.7)	20 (8.1)	3 (3.9)	5 (6.3)
25%-50%	24 (17.9)	2 (3.9)	6 (2.4)	1 (1.3)	0 (0)
≥ 50%	9 (6.8)	0 (0)	0 (0)	0 (0)	0 (0)
CA19-9
Negative	75 (56.0)	34 (66.7)	206 (83.5)	70 (90.8)	75 (93.8)
Positive
1%-10%	43 (32.1)	15 (29.4)	40 (16.1)	6 (7.9)	4 (5.0)
10%-25%	13 (9.7)	1 (2.0)	1 (0.4)	0 (0)	1 (1.3)
≥ 25%	3 (2.2)	1 (2.0)	0 (0)	1 (1.3)	0 (0)
CEA
Negative	34 (25.4)	20 (39.2)	189 (76.7)	59 (76.3)	67 (83.8)
Positive
1%-10%	47 (35.1)	27 (52.9)	56 (22.5)	16 (21.1)	7 (8.8)
10%-25%	31 (23.2)	4 (7.8)	2 (0.8)	1 (1.3)	6 (7.5)
25%-50%	19 (14.2)	0 (0)	0 (0)	1 (1.3)	0 (0)
≥ 50%	3 (2.2)	0 (0)	0 (0)	0 (0)	0 (0)
Three biomarkers combined2
Negative	21 (15.7)	12 (23.5)	129 (52.4)	44 (57.1)	58 (72.5)
Positive
1%-10%	35 (26.1)	26 (51.0)	90 (36.4)	28 (36.4)	13 (16.3)
10%-25%	30 (22.4)	10 (19.6)	22 (8.8)	4 (5.2)	9 (11.3)
25%-50%	36 (26.8)	3 (5.9)	6 (2.4)	1 (1.3)	0 (0)
≥ 50%	12 (9.0)	0 (0)	0 (0)	0 (0)	0 (0)

¹In the five groups, the Kruskal-Wallis H test values of the differences in positive protein expression ratios of p53, CA19-9, CEA, and the three biomarkers combined were 50.279 (P < 0.001), 68.660 (P < 0.001), 160.453 (P < 0.001), and 142.393 (P < 0.001), respectively. The values of the linear-by-linear association test were 68.737 (P < 0.001), 61.076 (P < 0.001), 128.738 (P < 0.001), and 141.591 (P < 0.001), respectively;

²If the expression of any one of the three biomarkers was positive, the combined expression was counted as positive; if there was an overlap in expression, the score of the strongest expression was taken as the score of the combination. ESCC: Esophageal squamous cell cancer; HGD: High-grade dysplasia; LGD: Low-grade dysplasia; BCH: Basal cell hyperplasia.

Table 3 The correlation coefficients for p53, CA19-9, CEA and selected variables

Variables	Spearman's correlation coefficient
	p53	CA19-9	CEA
Age	-0.0811	-0.065	-0.0931
School year	0.197	0.139	0.213
Alcohol drinking index	-0.050	-0.056	-0.021
Smoking index	-0.038	-0.050	-0.003
Family history of esophageal cancer	0.020	-0.078	-0.035
Degree of esophageal lesions3	0.2872	0.3262	0.4552
p53	1	0.3252	0.3742
CA19-9		1.000	0.5032
CEA			1.000

¹Correlation is significant at the 0.05 level (2-tailed);

²Correlation is significant at the 0.01 level (2-tailed);

³The variable definitions for the degree of esophageal lesions were: 1, normal control; 2, basal cell hyperplasia; 3, low-grade dysplasia; 4, high-grade dysplasia; 5, invasive squamous cell cancer.

Table 4 The associations of the positive expression of p53, CA19-9 and CEA proteins with basal cell hyperplasia, low-grade dysplasia, high-grade dysplasia, and esophageal squamous cell cancer [OR (95%CI)]

Variables	ESCC	HGD	LGD	BCH
p53
Model 11	4.51 (2.38-8.59)	4.50 (2.07-9.78)	2.21 (1.21-4.06)	1.50 (0.71-3.15)
Model 22	4.52 (2.17-4.92)	4.88 (2.10-11.33)	2.31 (1.17-4.56)	1.60 (0.72-3.57)
CA19-9
Model 11	15.15 (7.45-30.84)	7.99 (3.53-18.10)	1.57 (0.81-3.04)	1.60 (0.72-3.54)
Model 22	23.57 (9.27-56.86)	12.01 (4.61-31.32)	2.05 (0.91-4.63)	1.99 (0.81-4.93)
CEA
Model 11	11.80 (4.49-31.05)	7.50 (2.56-22.00)	2.96 (1.13-7.76)	1.52 (0.46-5.02)
Model 22	17.40 (4.76-63.59)	11.06 (2.80-43.71)	3.80 (1.06-13.66)	1.81 (0.41-7.96)
Three biomarkers combined3
Model 11	14.19 (7.21-27.91)	8.57 (3.80-19.30)	2.40 (1.38-4.15)	1.98 (1.02-3.85)
Model 22	16.64 (7.68-36.05)	10.11 (4.20-24.32)	2.59 (1.38-4.86)	2.15 (1.04-4.45)

¹In the regression model, ORs were calculated without adjustments for any variables;

²In the regression model, ORs were calculated after adjustment for age, school year, income per year-person, smoking, alcohol drinking, and family history of esophageal cancer;

³If the expression of any one of the three biomarkers was positive, the combined expression was counted as positive; if there was an overlap in expression, the score of the strongest expression was taken as the score of the combination. ESCC: Esophageal squamous cell cancer; HGD: High-grade dysplasia; LGD: Low-grade dysplasia; BCH: Basal cell hyperplasia.

Table 5 Area of receiver operating characteristic curves for p53, CA19-9, CEA and the three biomarkers combined in the four esophageal lesions

Protein expression	Area	SE	95%CI	Z value	P value
p53
ESCC	0.696	0.0357	0.629-0.756	5.483	0.001
HGD	0.669	0.0495	0.581-0.749	3.416	0.006
LGD	0.578	0.0356	0.523-0.632	2.197	0.028
BCH	0.533	0.0461	0.452-0.613	0.725	0.468
CA19-9
ESCC	0.650	0.0359	0.625-0.751	5.288	< 0.001
HGD	0.638	0.0509	0.549-7.200	2.714	0.007
LGD	0.554	0.0363	0.498-0.609	1.484	0.138
BCH	0.515	0.0464	0.433-0.595	0.317	0.751
CEA
ESCC	0.802	0.0293	0.742-0.853	10.302	< 0.001
HGD	0.712	0.0479	0.260-0.788	4.423	< 0.001
LGD	0.532	0.0369	0.476-0.558	0.872	0.383
BCH	0.531	0.0463	0.449-0.611	0.668	0.504
Three biomarkers combined1
ESCC	0.837	0.0260	0.780-0.884	212.656	< 0.001
HGD	0.740	0.0460	0.656-0.813	5.220	< 0.001
LGD	0.590	0.0352	0.535-0.644	2.586	0.010
BCH	0.562	0.0458	0.0481-0.641	1.360	0.174

¹If the expression of any one of the three biomarkers was positive, the combined expression was counted as positive; if there was an overlap in expression, the score of the strongest expression was taken as the score of the combination. ESCC: Esophageal squamous cell cancer; HGD: High-grade dysplasia; LGD: Low-grade dysplasia; BCH: Basal cell hyperplasia.

Table 6 Diagnostic values of predicting the four esophageal lesions based on the positive expression of the three biomarkers combined at the cut off value of 10%-25%¹ (%)

Group	Sensitivity	95%CI	+PV	95%CI
ESCC	58.2	49.4-66.7	89.7	81.3-95.1
HGD	25.5	14.3-39.6	59.1	36.4-79.3
LGD	11.2	7.6-15.8	75.7	58.8-88.2
BCH	6.5	2.2-14.5	35.7	12.9-64.8

¹If the expression of any one of the three biomarkers was positive, the combined expression was counted as positive; if there was an overlap in expression, the score of the strongest expression was taken as the score of the combination. The specificity was 88.8% for the normal control group, with a 95%CI of 79.7%-94.7%. ESCC: Esophageal squamous cell cancer; HGD: High-grade dysplasia; LGD: Low-grade dysplasia; BCH: Basal cell hyperplasia.