Patients basic information
In this study, 2259 ostensibly healthy individuals (990 males with ages from 14–85, and 1269 females with ages from 13–87), 1570 lung cancer patients (1055 males and 515 females, ages from 21–90), benign non-tumors (316 males and 223 females, ages from 14–92), and other malignancies (33 males and 22 females, ages from 22–90) were enrolled, as seen in Supplementary Table S1.
Serum ProGRP, NSE, SCC-Ag, CEA, CYFRA21-1, and HE4 levels in lung cancer, benign non-tumor diseases, and healthy controls
Serum ProGRP, NSE, SCC-Ag, CEA, CYFRA21-1, and HE4 levels were compared among lung cancers, benign non-cancerous diseases, and healthy controls (One-Way ANOVA). All 6 biomarkers in lung cancers were significantly higher than in benign non-cancerous disease and healthy controls (p < 0.0001 for all), while, apart from HE4, there was no significant difference in serum biomarkers levels between benign non-cancerous diseases and healthy controls (p > 0.05 for all). The HE4 level in benign non-cancerous diseases was significantly higher than in healthy controls (p = 0.000) and, further, there was no significant difference between lung cancer and benign non-cancerous diseases (p = 0.140) (Fig. 1).
Comparison of serum ProGRP, NSE, SCC-Ag, CEA, CYFRA21-1, and HE4 levels between lung cancer and other types of malignancies
As shown in Table 1, serum SCC-Ag, CEA, and CYFRA21-1 levels in lung cancer patients are significantly higher than in other types of malignancies (p = 0.000, 0.000, and 0.002, respectively). There is no significant difference in serum ProGRP, NSE, and H4 levels between lung cancers and other types of malignancies (p = 0.495, 0.328, and 0.844, respectively).
Table 1
Comparison of 6 lung cancer biomarkers between lung cancer and other types of malignancies (independent-samples t-test)
|
Diagnosis
|
n
|
Average
|
STDEV
|
t
|
p
|
ProGRP
|
Lung cancer
|
1570
|
207.67
|
776.86
|
0.682
|
0.495
|
|
Other types of malignancies
|
55
|
135.27
|
668.40
|
|
|
NSE
|
Lung cancer
|
1570
|
25.79
|
54.43
|
0.979
|
0.328
|
|
Other types of malignancies
|
55
|
18.58
|
18.97
|
|
|
SCC
|
Lung cancer
|
1570
|
2.53
|
7.17
|
5.461
|
0.000
|
|
Other types of malignancies
|
55
|
1.25
|
1.11
|
|
|
CEA
|
Lung cancer
|
1570
|
31.77
|
127.82
|
9.016
|
0.000
|
|
Other types of malignancies
|
55
|
2.54
|
2.30
|
|
|
CYFRA21-1
|
Lung cancer
|
1570
|
10.87
|
26.44
|
3.184
|
0.002
|
|
Other types of malignancies
|
55
|
5.67
|
11.05
|
|
|
HE4
|
Lung cancer
|
1570
|
119.68
|
152.65
|
-0.197
|
0.844
|
|
Other types of malignancies
|
54
|
123.83
|
131.08
|
|
|
Correlation of serum ProGRP, NSE, SCC-Ag, CEA, CYFRA21-1, and HE4 levels with histopathology
As seen in Fig. 2, ProGRP and NSE levels in SCLC are significantly higher than in other histologic types (p = 0.000 for all); SCC-Ag and CYFRA21-1 levels in SCC are significantly higher than in other histologic types (p = 0.000 for all); the HE4 level in SCLC is also significantly higher than in SCC and adenocarcinoma (p = 0.004 and p = 0.028, respectively); while adenocarcinoma has the highest serum CEA level comparing to SCC and SCLC (p = 0.000 and p = 0.066, respectively).
Comparison of serum ProGRP, NSE, SCC-Ag, CEA, CYFRA21-1, and HE4 levels between SCLC and NSCLC
Table 2 indicates that serum ProGRP, NSE, and HE4 in SCLC are significantly higher than in NSCLC (p = 0.000, 0.000, and 0.009, respectively), while serum SCC-Ag level in NSCLC is significantly higher than in SCLC (p = 0.000). There is no significant difference in serum CEA and CYFRA21-1 between SCLC and NSCLC (p = 0.366 and 0.146, respectively).
Table 2
Comparison of lung cancer biomarkers between small cell lung cancer and non-small cell lung cancer (independent-samples t-test)
|
Histopathology
|
n
|
Mean
|
STDEV
|
t
|
p
|
ProGRP
|
NSCLC
|
1316
|
49.24
|
247.36
|
-10.131
|
0.000
|
|
SCLC
|
196
|
1295.75
|
1719.92
|
|
|
NSE
|
NSCLC
|
1316
|
17.48
|
31.18
|
-7.789
|
0.000
|
|
SCLC
|
196
|
81.89
|
115.12
|
|
|
SCC
|
NSCLC
|
1315
|
2.68
|
7.52
|
5.718
|
0.000
|
|
SCLC
|
196
|
1.24
|
1.99
|
|
|
CEA
|
NSCLC
|
1316
|
32.17
|
127.93
|
0.905
|
0.366
|
|
SCLC
|
196
|
23.48
|
106.90
|
|
|
CYFRA21-1
|
NSCLC
|
1316
|
11.03
|
24.25
|
1.453
|
0.146
|
|
SCLC
|
196
|
8.14
|
35.24
|
|
|
HE4
|
NSCLC
|
1316
|
114.25
|
145.48
|
-2.602
|
0.009
|
|
SCLC
|
196
|
144.45
|
187.83
|
|
|
Correlation Of 6 Lung Cancer Biomarkers With Clinical Stages
As shown in Table 3, serum levels for all 6 biomarkers in advanced clinical stages (III + IV) are significantly higher than in low stages (I + II) (p = 0.000, 0.000, 0.001, 0.000, 0.000, and 0.000, respectively).
Table 3
Correlation of lung cancer biomarkers with clinical stages (independent-samples t-test)
|
Clinical stages
|
n
|
Mean
|
STDEV
|
t
|
P value
|
ProGRP
|
I + II
|
428
|
49.10
|
103.49
|
-7.946
|
0.000
|
|
III + IV
|
1004
|
292.22
|
956.47
|
|
|
NSE
|
I + II
|
428
|
16.98
|
51.62
|
-4.479
|
0.000
|
|
III + IV
|
1004
|
30.88
|
58.42
|
|
|
SCC
|
I + II
|
427
|
1.73
|
4.89
|
-3.231
|
0.001
|
|
III + IV
|
1004
|
2.84
|
7.80
|
|
|
CEA
|
I + II
|
428
|
5.26
|
37.02
|
-7.789
|
0.000
|
|
III + IV
|
1004
|
45.97
|
155.58
|
|
|
CYFRA21-1
|
I + II
|
428
|
4.96
|
15.20
|
-7.215
|
0.000
|
|
III + IV
|
1004
|
13.71
|
30.580
|
|
|
HE4
|
I + II
|
428
|
65.68
|
55.70
|
-12.685
|
0.000
|
|
III + IV
|
1004
|
144.95
|
178.69
|
|
|
Diagnostic performance of 6 lung cancer biomarkers on histological types of lung cancers
Table 4 shows the diagnostic performance of a single biomarker on a specific histologic type of lung cancer. The results indicate that for the SCLC histologic type, NSE, HE4, and ProGRP present relative high performance (AUC: 0.881, 0.856, and 0.834, respectively; sensitivity: 76.14, 83.76, and 70.05, respectively; specificity: 89.16, 78.53, and 97.78, respectively), with NSE presenting the highest AUC, HE4 showing the highest sensitivity, and ProGRP having the highest specificity. For SCC histologic type, CYFRA21-1, HE4, and SCC-Ag indicate relative high performance (AUC: 0.920, 0.832, and 0.793, respectively; sensitivity: 81.02, 76.64, and 59.12, respectively; specificity: 89.13, 79.57, and 91.02, respectively), with CYFRA21-1 showing the highest AUC and sensitivity, SCC-Ag having the highest specificity. For the adenocarcinoma histologic type, the performance of all biomarkers were generally low, HE4, CYFRA21-1, and CEA had relative high performance (AUC: 0.732, 0.722, and 0.708, respectively; sensitivity: 64.99, 66.34, and 42.79, respectively; specificity: 71.52, 85.69, and 91.20, respectively), with HE4 presenting the highest AUC, CYFRA21-1 showing the highest sensitivity, and CEA having the highest specificity. For NSCLC as a pooled histologic type, the performance of all biomarkers were also low, CYFRA21-1, HE4, CEA, and SCC-Ag have relative high performance (AUC: 0.787, 0.765, 0.712, and 0.610, respectively; sensitivity: 62.46, 66.03, 40.43and 30.70, respectively; specificity: 81.72, 76.60, 91.20, and 90.88, respectively), with CEA and SCC-Ag having the highest specificity.
Table 4
Diagnostic performance of 6 lung cancer biomarkers on histological types of lung cancers
Histopathology
|
Biomarker
|
AUC
|
Youden’s index
|
Cut-off
|
Sensitivity
|
Specificity
|
SCLC
|
ProGRP
|
0.834
|
0.6783
|
78.71
|
70.05
|
97.78
|
|
NSE
|
0.881
|
0.6531
|
19.40
|
76.14
|
89.16
|
|
SCC
|
0.544
|
0.1208
|
0.86
|
54.31
|
57.76
|
|
CEA
|
0.763
|
0.3889
|
1.63
|
82.74
|
56.74
|
|
CYFRA21-1
|
0.808
|
0.4724
|
2.37
|
76.14
|
71.10
|
|
HE4
|
0.856
|
0.6229
|
62.55
|
83.76
|
78.53
|
SCC
|
ProGRP
|
0.602
|
0.2353
|
23.31
|
34.79
|
88.73
|
|
NSE
|
0.586
|
0.1773
|
15.63
|
42.34
|
75.39
|
|
SCC
|
0.793
|
0.5015
|
1.48
|
59.12
|
91.02
|
|
CEA
|
0.715
|
0.2999
|
1.96
|
63.50
|
66.49
|
|
CYFRA21-1
|
0.920
|
0.7015
|
3.25
|
81.02
|
89.13
|
|
HE4
|
0.832
|
0.5621
|
63.68
|
76.64
|
79.57
|
Adeno-C
|
ProGRP
|
0.570
|
0.1174
|
31.12
|
41.53
|
70.21
|
|
NSE
|
0.555
|
0.1163
|
16.42
|
32.27
|
79.36
|
|
SCC
|
0.525
|
0.0890
|
1.33
|
21.74
|
87.16
|
|
CEA
|
0.708
|
0.3399
|
3.48
|
42.79
|
91.20
|
|
CYFRA21-1
|
0.722
|
0.3203
|
2.98
|
66.34
|
85.69
|
|
HE4
|
0.732
|
0.3651
|
56.12
|
64.99
|
71.52
|
NSCLC
|
ProGRP
|
0.576
|
0.1447
|
23.9
|
26.60
|
87.88
|
|
NSE
|
0.567
|
0.1378
|
15.88
|
36.78
|
77.00
|
|
SCC
|
0.610
|
0.2158
|
1.47
|
30.70
|
90.88
|
|
CEA
|
0.712
|
0.3163
|
3.48
|
40.43
|
91.20
|
|
CYFRA21-1
|
0.787
|
0.4419
|
2.77
|
62.46
|
81.72
|
|
HE4
|
0.765
|
0.4263
|
60.37
|
66.03
|
76.60
|
Diagnostic performance of combinatory analysis of 6 biomarkers on histological types of lung cancers
Different combinatory analyses were performed based on the diagnostic performance of biomarkers from previous reports and from the single biomarker’s performance in this study. Table 5 shows the diagnostic performance of combinatory analysis of 6 biomarkers on histologic types of lung cancers. For a SCC histologic type, the best performance of a 2-marker combinatory pattern is from SCC-Ag + CYFRA21-1 with AUC, sensitivity, and specificity as 0.923, 81.51, and 91.49, respectively. A combination of 3 biomarkers (SCC-Ag + CEA + CYFRA21-1) shows slight improvement on diagnostic performance, with AUC and specificity increased, but the sensitivity, however, reduced slightly. For SCLC, the best performance of 2-marker combinatory pattern is from ProGRP + NSE with AUC, sensitivity, and specificity as 0.928, 82.74, and 97.53, respectively. A three-marker combinatory pattern (ProGRP + NSE + HE4) slightly increased the AUC and sensitivity, but the specificity was reduced. For an adenocarcinoma histologic type, the diagnostic performance is generally low, the 2-marker combination (CEA + CYFRA21-1) and 3-marker combination (CEA + CYFRA21-1 + HE4) gained similar diagnostic performance with AUC, sensitivity, and specificity as 0.752, 50.46, and 90.41, respectively, and 0.752, 50.46, and 90.48, respectively. For NSCLC, the highest performance of 2-marker combination is CEA + CYFRA21-1, with AUC, sensitivity, and specificity as 0.809, 60.26, and 89.74. The highest performance for 3-marker combination is SCC + CEA + CYFRA21-1, with AUC, sensitivity, and specificity as 0.810, 61.17, 90.13, respectively, which gained highest diagnostic performance (Table 5).
Table 5
Diagnostic performance of combinatory analysis of 6 lung cancer biomarkers on histological types of lung cancers
|
Combination
|
AUC
|
Youden index
|
Sensitivity
|
Specificity
|
SCC
|
SCC + CYFRA21-1
|
0.923
|
0.7300
|
81.51
|
91.49
|
|
CEA + CYFRA21-1
|
0.920
|
0.7099
|
82.97
|
88.02
|
|
HE4 + CYFRA21-1
|
0.914
|
0.6914
|
78.81
|
89.34
|
|
SCC + CEA + CYFRA21-1
|
0.924
|
0.7287
|
81.02
|
91.85
|
|
SCC + HE4 + CYFRA21-1
|
0.917
|
0.7185
|
76.64
|
95.21
|
|
CEA + HE4 + CYFRA21-1
|
0.911
|
0.6967
|
80.29
|
89.37
|
SCLC
|
ProGRP + NSE
|
0.928
|
0.8027
|
82.74
|
97.53
|
|
NSE + HE4
|
0.889
|
0.6605
|
76.14
|
89.91
|
|
NSE + CYFRA21-1
|
0.886
|
0.6556
|
76.65
|
88.91
|
|
CYFRA21-1 + HE4
|
0.872
|
0.6159
|
77.16
|
84.44
|
|
ProGRP + HE4
|
0.859
|
0.6899
|
73.60
|
95.38
|
|
ProGRP + NSE + HE4
|
0.929
|
0.8036
|
83.76
|
96.60
|
Adeno-C
|
CEA + CYFRA21-1
|
0.752
|
0.4087
|
50.46
|
90.41
|
|
CYFRA21-1 + HE4
|
0.738
|
0.3590
|
54.58
|
81.32
|
|
CEA + HE4
|
0.729
|
0.3624
|
46.80
|
89.45
|
|
CEA + CYFRA21-1 + HE4
|
0.752
|
0.4094
|
50.46
|
90.48
|
NSCLC
|
CEA + CYFRA21-1
|
0.809
|
0.4999
|
60.26
|
89.74
|
|
CYFRA21-1 + HE4
|
0.796
|
0.4636
|
59.95
|
86.40
|
|
CEA + HE4
|
0.755
|
0.3721
|
58.97
|
78.25
|
|
CEA + CYFRA21-1 + HE4
|
0.807
|
0.4960
|
62.77
|
86.83
|
|
SCC + CEA + CYFRA21-1
|
0.810
|
0.5130
|
61.17
|
90.13
|
|
SCC + HE4 + CYFRA21-1
|
0.796
|
0.4694
|
62.01
|
84.94
|