Cohort characteristic
Between January 2015 and September 2020, a total of 384 patients underwent hepatectomy and had histologically confirmed HCC at the Affiliated Lihuili Hospital of Ningbo University. Excluded from this cohort were 177 patients without regularly monitoring the three biomarkers every 3 months after the operation, 12 patients with other malignancies, 3 patients with a positive resection margin, and 4 patients with long-term use of vitamin K antagonists (Figure 1). After exclusion, all 188 patients with more than 6 months of follow-up were eligible for further analysis. The clinicopathological characteristics of these patients are summarized, there were significant differences in maximum tumor size, tumor number, portal vein tumor thrombus, microvascular invasion, tumor capsule, and TNM stage between the recurrence and no recurrence patients (Table 1).
TABLE 1: Clinicopathological characteristics of the recurrence patients and no recurrence patients
Characteristics
|
Recurrence
(n=69)
|
No recurrence
(n=119)
|
χ2/Z
|
P
|
Age (years,%)
≥60
<60
|
33(47.8)
36(52.2)
|
59(49.6)
60(50.4)
|
0.054
|
0.817
|
Sex (%)
Male
Female
|
57(82.6)
12(17.4)
|
89(74.8)
30(25.2)
|
1.539
|
0.215
|
HBsAg (%)
Positive
Negative
|
56(81.2)
13(18.8)
|
92(77.3)
27(22.7)
|
0.386
|
0.534
|
ALB (g/L, %)
≥40
<40
|
41(59.4)
28(40.6)
|
75(63.0)
44(37.0)
|
0.240
|
0.624
|
ALT (U/L, %)
≥40
<40
|
17(24.6)
52(75.4)
|
33(27.7)
86(72.3)
|
0.214
|
0.644
|
Maximum tumor size (cm, %)
≥6
<6
|
25(36.2)
44(63.8)
|
25(21.0)
94(79.0)
|
5.185
|
0.023
|
Number of tumors (%)
Single
Multiple
|
47(68.1)
22(31.9)
|
97(81.5)
22(18.5)
|
4.373
|
0.037
|
Differentiation (%)
Poor
Well-moderate
|
32(46.4)
37(53.6)
|
42(35.3)
77(64.7)
|
2.248
|
0.134
|
Portal vein tumor thrombosis (%)
Present
Absent
|
12(17.4)
57(82.6)
|
6(5.0)
113(95.0)
|
7.693
|
0.006
|
Microvascular invasion (%)
Present
Absent
|
39(56.5)
30(43.5)
|
46(38.7)
73(61.3)
|
5.628
|
0.018
|
Perineural invasion (%)
Present
Absent
|
2(2.9)
67(97.1)
|
2(1.7)
117(98.3)
|
0.311
|
0.577
|
Tumor capsule (%)
Present
Absent
|
48(69.6)
21(30.4)
|
99(83.2)
20(16.8)
|
4.757
|
0.029
|
Child-Pugh class (%)
A
B
|
67(97.1)
2(2.9)
|
117(98.3)
2(1.7)
|
0.311
|
0.577
|
TNM stage (%)
III-IV
I-II
|
45(65.2)
24(34.8)
|
103(86.6)
16(13.4)
|
11.872
|
0.001
|
PIVKA-Ⅱ (mAU/ml)
Month -6
Month -3
Month 0
|
23(14,110)
59(20,286)
161(27,666)
|
14(12,19)
15(12,18)
15(12,19)
|
4.677
8.324
8.318
|
<0.001
<0.001
<0.001
|
AFP (ng/L)
Month -6
Month -3
Month 0
|
6.2(2.4,26.2)
9.2(2.7,68.3)
14.2(2.5,164.7)
|
2.8(1.8,4.8)
2.9(1.7,4.8)
2.6(1.5,5.1)
|
4.885
5.347
5.455
|
<0.001
<0.001
<0.001
|
AFP-L3 (%)
Month -6
Month -3
Month 0
|
0.5(0.5,10.4)
1.7(0.5,34.5)
2.4(0.5,39.1)
|
0.5(0.5,0.5)
0.5(0.5,0.5)
0.5(0.5,0.5)
|
3.348
6.912
8.111
|
0.001
<0.001
<0.001
|
ALB: albumin; ALT: Alanine transaminase.
Data are presented as number (percentages). PIVKA-II, AFP, AFP-L3 are presented as median (25th quantile, 75th quantile).
Trends in PIVKA-Ⅱ, AFP and AFP-L3
In the GEE analysis, PIVKA-II, AFP, AFP-L3 in the recurrence patients were significantly higher than the no recurrence patients from month -6 to month 0 (P≤0.001, Table 1). PIVKA-Ⅱ and AFP showed increasing trends from month -6 to month 0 in the recurrence patients, and there were significant differences compared with the trends in the no recurrence patients (P=0.001, P<0.001 respectively), but AFP-L3 had no such difference (P=0.39, Figure 2, Table 2). These indicate that PIVKA-Ⅱ and AFP effects are different according to the time period.
TABLE 2: Comparison the trends of three biomarkers between recurrence and no recurrence patients
|
Estimate
|
Standard error
|
Z
|
P
|
PIVKA-Ⅱ
|
Group(mAU/ml)
|
1735.3
|
797.6
|
2.18
|
0.030
|
Group * time
(month -6 vs -3)
|
964.9
|
642.3
|
1.50
|
0.133
|
Group * time
(month -6 vs 0)
|
3713.0
|
1779.7
|
2.09
|
0.04
|
AFP
|
Group (ng/L)
|
206.7
|
60.6
|
3.41
|
<0.001
|
Group * time
(month -6 vs -3)
|
118.5
|
47.7
|
2.49
|
0.012
|
Group * time
(month -6 vs 0)
|
368.6
|
110.9
|
3.32
|
<0.001
|
AFP-L3
|
Group (%)
|
9.8
|
3.5
|
16.68
|
0.005
|
Group * time
(month -6 vs -3)
|
7.5
|
4.8
|
1.55
|
0.121
|
Group * time
(month -6 vs 0)
|
4.7
|
5.5
|
0.86
|
0.39
|
Performance to predict recurrence
In the performance of predict recurrence, the AUROCs for PIVKA-Ⅱ, AFP, AFP-L3 at month 0 were 0.885, 0.754, 0.781 respectively; The AUROCs for PIVKA-Ⅱ, AFP, AFP-L3 at month -3 were 0.871, 0.748, 0.744 respectively; The AUROCs for PIVKA-Ⅱ, AFP, AFP-L3 at month -6 were 0.718, 0.708, 0.603 respectively. The combination of the three biomarkers can improve the performance to predict recurrence, the AUROCs at month -6, month -3, month 0 was 0.786, 0.895, and 0.885 respectively (Figure 3, Table 3).
TABLE 3: AUROC for PIVKA- Ⅱ, AFP, AFP-L3 and combinations in predicting recurrence
Month from recurrence
|
Month -6 AUROC (95%CI)
|
Month -3 AUROC (95%CI)
|
Month 0 AUROC (95%CI)
|
PIVKA-Ⅱ
|
0.718(0.633-0.803)
|
0.871(0.813-0.930)
|
0.885(0.827-0.943)
|
AFP
|
0.708(0.621-0.795)
|
0.748(0.669-0.827)
|
0.754(0.675-0.833)
|
AFP-L3
|
0.603(0.512-0.693)
|
0.744(0.664-0.825)
|
0.781(0.703-0.858)
|
PIVKA-Ⅱ+AFP
|
0.789(0.714-0.864)
|
0.886(0.830-0.942)
|
0.884(0.823-0.944)
|
PIVKA-Ⅱ+AFPL3
|
0.733(0.649-0.817)
|
0.891(0.836-0.946)
|
0.888(0.831-0.946)
|
AFP+AFPL3
|
0.701(0.615-0.786)
|
0.772(0.694-0.850)
|
0.782(0.705-0.860)
|
PIVKA-Ⅱ+AFP+AFPL3
|
0.786(0.711-0.862)
|
0.895(0.840-0.950)
|
0.885(0.824-0.944)
|
Optimal cut-off value and the performance
At month 0, the optimal cut-off value to predict recurrence for PIVKA-Ⅱ, AFP, AFP-L3 were 29.5mAU/ml, 10.7ng/L, 1.5% respectively. With the optimal cut-off value, the sensitivity in predicting recurrence for PIVKA-Ⅱ, AFP, AFP-L3 at month -6 were 42.2%, 37.5%, 34.4% respectively; the sensitivity at month -3 were 68.8%, 50%, 51.6% respectively; the sensitivity at month 0 were 75.0%, 54.7%, 57.8% respectively. The combination of the three biomarkers can improve the performance, and the sensitivities at month -6, month -3, month 0 were 60.9%, 79.7%, 79.7% respectively (Table 4).
TABLE 4: Performance of three biomarkers and combinations in predicting recurrence
Month from recurrence
|
Sensitivity(%)
|
Specificity(%)
|
PIVKA-Ⅱ≥29.5mAU/ml
|
|
|
Month 0
|
75.0%
|
94.7%
|
Month -3
|
68.8%
|
96.8%
|
Month -6
|
42.2%
|
91.6%
|
AFP≥10.7ng/L
|
|
|
Month 0
|
54.7%
|
96.8%
|
Month -3
|
50%
|
96.8%
|
Month -6
|
37.5%
|
95.8%
|
AFPL3≥1.5%
|
|
|
Month 0
|
57.8%
|
96.8%
|
Month -3
|
51.6%
|
92.6%
|
Month -6
|
34.4%
|
86.3%
|
PIVKA-Ⅱ≥29.5mAU/ml +AFP≥10.6ng/L
|
|
|
Month 0
|
79.7%
|
94.5%
|
Month -3
|
76.8%
|
94.7%
|
Month -6
|
54.7%
|
91.6%
|
PIVKA-Ⅱ≥29.5mAU/ml +AFPL3≥1.5%
|
|
|
Month 0
|
78.3%
|
93.6%
|
Month -3
|
76.8%
|
90.4%
|
Month -6
|
51.6%
|
83.2%
|
AFP≥10.7ng/L +AFPL3≥1.5%
|
|
|
Month 0
|
65.2%
|
96.4%
|
Month -3
|
58.0%
|
91.2%
|
Month -6
|
43.8%
|
84.1%
|
PIVKA-Ⅱ>29.5mAU/ml+AFP>10.7ng/L
+AFPL3>1.5%
|
|
|
Month 0
|
79.7%
|
93.6%
|
Month -3
|
79.7%
|
89.5%
|
Month -6
|
60.9%
|
83.2%
|
Elevated biomarkers correlate to recurrence
The median RFS of all patients was 22 months (n=188, 40 months, 95% CI 26.539-53.461), and the median RFS of patients with any biomarkers elevated above the optimal cut-off value during the follow-up (n=82, 19 months, 95% CI 14.757-23.243) was significantly shorter than that of patients without elevated biomarker (n=106, 58 months, 95% CI 51.575-65.022) (χ2=62.125, P<0.001, Figure 4).