The expression of hepatocellular carcinoma (HCC) tumor markers, alpha-fetoprotein (AFP), and protein induced by vitamin K absence or antagonist-II (PIVKA-II or des-gamma-carboxy prothrombin) is highly variable in patients with HCC [1,2]. These tumor markers are also variably expressed even in patients without HCC, especially in patients with liver cirrhosis. As a result, the threshold of AFP and PIVKA-II for diagnosis of HCC in patients with liver cirrhosis is much higher than in patients with normal or non-cirrhotic livers [3]. Among the two HCC biomarkers, AFP is more frequently emphasized than PIVKA-II worldwide including in Korea [4-6].

It has been suggested that there is a possibility of different expression of HCC tumor markers according to the background liver diseases. This study was intended to compare the expression patterns and diagnostic reliability of serum AFP and PIVKA-II in patients with or without HCC who were undergoing living donor liver transplantation (LDLT).

This was a retrospective single-center observational study on the expression of HCC tumor markers in LDLT candidates. The liver transplantation (LT) database of Asan Medical Center was searched to identify adult patients who underwent primary LDLT with or without HCC during a 13-year period from January 2006 to June 2018. The inclusion criterion was age ≥18 years at the time of LT and primary LDLT. Patients who did not measure AFP and PIVKA-II within 2 weeks of the LT operation were excluded. We finally selected 3,091 patients for this study. We divided the LT recipients according to the diagnosis of HCC and background liver diseases (hepatitis B virus [HBV]-associated liver cirrhosis, hepatitis C virus [HCV]-associated liver cirrhosis, alcoholic liver disease [ALD], and others). Diagnosis of HCC was determined according to the explant pathology reports. The institutional review board of the Asan Medical Center approved this study protocol, which waived the requirement for informed consent due to the retrospective nature of this study. This study was performed in accordance with the ethical guidelines of the World Medical Association Declaration of Helsinki 2013.

All numerical data are presented as either the means with standard deviations or as the medians. Continuous variables were compared using the analysis of variance (ANOVA), and incidence variables were compared using the chi-squared test. The cutoff of AFP and PIVKA-II for predicting HCC was determined by using receiver operating characteristic (ROC) curve analysis. The optimal cutoff, sensitivity, and specificity were determined using the Youden index. Differences with p-values <0.05 were considered statistically significant. All statistical analyses were performed using SPSS version 22 (IBM Corp., Armonk, NY, USA) and MedCal version 20.010 (MedCal, Ostend, Belgium).

Profiles of All the Patients

The number of patients with and without HCC were 1,595 and 1,496, respectively. The number of patients with and without HCC were 1,297 and 719 in the HBV group; 128 and 90 in the HCV group; 109 and 381 in the ALD group; and 61 and 306 in other disease group, respectively.

The mean and median AFP level were 194.1±2,068.9 ng/mL and 9.0 ng/mL in the HCC patients and 26.8 ±150.2 ng/mL and 4.0 ng/mL in the non-HCC patients, respectively (p<0.001; Fig. 1). The mean and median PIVKA-II level was 279.7±2,406.2 mAU/mL and 25 mAU/mL in the HCC patients, and 205.9±1,033.2 ng/mL and 23 mAU/mL in the non-HCC patients, respectively (p=0.274; Fig. 2).

Figure 1. Comparison of alpha-fetoprotein (AFP) (A) and protein induced by vitamin K absence or antagonist-II (PIVKA-II) (B) levels according to the presence and absence of hepatocellular carcinoma (HCC) in all 3,091 patients. Bars indicate 25–75 percentiles.

Figure 2. Scatter plots on distribution of alpha-fetoprotein (AFP) and vitamin K absence or antagonist-II (PIVKA-II) values in patients with (A) and without (B) hepatocellular carcinoma (HCC).

ROC curve analyses for AFP showed that the area under the ROC curve (AUC) was 0.663; Youden index J was 0.233; the cutoff was 6.7 ng/mL with a sensitivity of 57.2% and specificity of 66.1% (Fig. 3). With the application of an institutional cutoff of 7.5 ng/mL, high AFP was observed in 857 of 1,595 (53.7%) in the HCC group and 470 of 1,496 (31.4%) the non-HCC patients (p<0.001), showing a sensitivity of 53.7%, specificity of 68.6%, positive predictive value (PPV) of 64.6%, negative predictive value (NPV) of 58.2%, and an accuracy of 60.9% (Table 1).

Table 1 .. Comparison of diagnostic predictability of hepatocellular carcinoma tumor markers.

Group	Cutoff
	AFP >7.5 ng/mL						PIVKA-II >40 mAU/mL						AFP >7.5 ng/mL or PIVKA-II >40 mAU/mL
	Sensitivity	Specificity	PPV	NPV	Accuracy		Sensitivity	Specificity	PPV	NPV	Accuracy		Sensitivity	Specificity	PPV	NPV	Accuracy
All	53.7	68.6	64.6	58.2	60.9		31.9	71.2	55.5	48.1	50.4		65.7	45.9	56.4	55.7	56.1
HBV	53.8	61.2	72.0	42.7	56.8		28.4	81.8	73.5	38.8	47.4		63.4	52.0	70.4	44.1	59.3
HCV	53.8	62.2	62.6	53.3	57.7		44.5	70.0	67.9	47.0	55.1		76.7	41.1	65.8	58.7	63.7
ALD	51.7	79.5	37.1	87.6	74.3		54.1	44.9	21.9	77.4	46.9		73.4	38.3	25.4	83.4	46.1
Other	55.7	71.9	28.3	89.1	69.2		41.0	60.8	17.2	83.8	57.5		72.1	42.5	20.0	88.4	47.4

Values are presented as percentage..

AFP, alpha-fetoprotein; PIVKA-II, protein induced by vitamin K absence or antagonist-II; HBV, hepatitis B virus; HCV, hepatitis C virus; ALD, alcoholic liver disease; PPV, positive predictive value; NPV, negative predictive value..

Figure 3. Receiver operating characteristic curve analysis of alpha-fetoprotein (AFP) and protein induced by vitamin K absence or antagonist-II (PIVKA-II) for diagnosis of hepatocellular carcinoma.

ROC curve analysis for PIVKA-II showed that the AUC was 0.536; Youden index J was 0.101; the cutoff was 15 mAU/mL with a sensitivity of 79.6% and specificity of 30.6% (Fig. 3). With the application of an institutional cutoff of 40 mAU/mL, high PIVKA-II was observed in 509 of 1,595 (31.9%) in the HCC group and 408 of 1,496 (27.3%) in the non-HCC patients (p=0.065), showing a sensitivity of 31.9%, specificity of 71.2%, PPV of 55.5%, NPV of 48.1%, and an accuracy of 50.4% (Table 1).

Profiles of Patients with HBV-Associated Liver Cirrhosis

The mean and median AFP level were 375.2±3,035.7 ng/mL and 9.0 ng/mL respectively in the HCC patients, and 31.3±162.7 ng/mL and 4.6 ng/mL respectively in the non-HCC patients (p=0.002; Fig. 4A). The mean and median PIVKA-II levels was 282.5±2,611.1 mAU/mL and 24 mAU/mL respectively in the HCC patients, and 85.0±532.0 ng/mL and 17 mAU/mL respectively in the non-HCC patients (p=0.045; Fig. 4B).

Figure 4. Comparison of alpha-fetoprotein (AFP) (A) and protein induced by vitamin K absence or antagonist-II (PIVKA-II) (B) levels according to the presence and absence of hepatocellular carcinoma (HCC) in patients with hepatitis B virus-associated liver cirrhosis. Bars indicate 25–75 percentiles. (C) Receiver operating characteristic curve analysis of AFP and PIVKA-II for diagnosis of HCC.

ROC curve analyses for AFP showed that the AUC was 0.634; Youden index J was 0.173; the cutoff was 4.3 ng/mL with a sensitivity of 68.6% and specificity of 48.7% (Fig. 4C). With the application of an institutional cutoff of 7.5 ng/mL, high AFP was observed in 698 of 1,297 (53.8%) in the HCC group, and 272 of 719 (37.8%) in the non-HCC group (p<0.001), showing a sensitivity of 53.8%, specificity of 61.2%, PPV of 72.0%, NPV of 42.7%, and an accuracy of 56.8% (Table 1).

ROC curve analyses for PIVKA-II showed that the AUC was 0.536; Youden index J was 0.101; the cutoff was 15 mAU/mL with a sensitivity of 79.6% and specificity of 30.6% (Fig. 4C). With the application of an institutional cutoff of 40 mAU/mL, high PIVKA-II was observed in 368 of 1,297 (28.4%) in the HCC group, and 131 of 719 (18.2%) in the non-HCC group (p<0.001), showing a sensitivity of 28.4%, specificity of 81.8%, PPV of 73.5%, NPV of 38.8%, and an accuracy of 47.4% (Table 1).

Profiles for Patients with HCV-Associated Liver Cirrhosis

The mean and median AFP level were 430.8±2,962.9 ng/mL and 9.0 ng/mL respectively in the HCC group and 13.8±22.8 ng/mL and 5.3 ng/mL respectively in the non-HCC group (p=0.184; Fig. 5A). The mean and median PIVKA-II level were 305.8±1,552.3 mAU/mL and 24 mAU/mL respectively in HCC group, and 96.2±474.8 ng/mL and 24 mAU/mL respectively in the non-HCC group (p=0.216; Fig. 5B).

Figure 5. Comparison of alpha-fetoprotein (AFP) (A) and protein induced by vitamin K absence or antagonist-II (PIVKA-II) (B) levels according to the presence and absence of hepatocellular carcinoma (HCC) in patients with hepatitis C virus-associated liver cirrhosis. Bars indicate 25–75 percentiles. (C) Receiver operating characteristic curve analysis of AFP and PIVKA-II for diagnosis of HCC.

ROC curve analyses for AFP showed that the AUC was 0.660; Youden index J was 0.275; the cutoff was 6.9 ng/mL with a sensitivity of 66.4% and specificity of 61.1% (Fig. 5C). With the application of an institutional cutoff of 7.5 ng/mL, high AFP was observed in 79 of 128 (61.7%) in the HCC group, and 34 of 90 (37.8%) in the non-HCC patients (p=0.026), showing a sensitivity of 53.8%, specificity of 62.2%, PPV of 62.6%, NPV of 53.3%, and an accuracy of 57.7% (Table 1).

ROC curve analyses for PIVKA-II showed that the AUC was 0.594; Youden index J was 0.156; the cutoff was 34 mAU/mL with a sensitivity of 50.0% and specificity of 65.6% (Fig. 5C). With the application of an institutional cutoff of 40 mAU/mL, high PIVKA-II was observed in 57 of 128 (44.5%) in the HCC group, and 27 of 90 (30.0%) in the non-HCC group (p=0.030), showing a sensitivity of 44.5%, specificity of 70.0%, PPV of 67.9%, NPV of 47.0%, and an accuracy of 55.1% (Table 1).

Profiles for Patients with ALD

The mean and median AFP level were 30.7±84.0 ng/mL and 6.2 ng/mL respectively in the HCC group, and 6.4±10.1 ng/mL and 3.6 ng/mL respectively in the non-HCC group (p<0.001; Fig. 6A). The mean and median PIVKA-II level were 279.2±855.6 mAU/mL and 54 mAU/mL respectively in the HCC group, and 246.0±552.1 ng/mL and 48 mAU/mL respectively in the non-HCC group (p=0.456; Fig. 6B).

Figure 6. Comparison of alpha-fetoprotein (AFP) (A) and protein induced by vitamin K absence or antagonist-II (PIVKA-II) (B) levels according to the presence and absence of hepatocellular carcinoma (HCC) in patients with alcoholic liver disease. Bars indicate 25–75 percentiles. (C) Receiver operating characteristic curve analysis of AFP and PIVKA-II for diagnosis of HCC.

ROC curve analyses for AFP showed that the AUC was 0.657; Youden index J was 0.268; the cutoff was 4.1 ng/mL with a sensitivity of 70.6% and specificity of 56.2% (Fig. 6C). With the application of an institutional cutoff of 7.5 ng/mL, a high AFP was observed in 46 of 109 (42.2%) in the HCC group, and 78 of 381 (20.5%) in the non-HCC group (p<0.001), showing a sensitivity of 51.7%, specificity of 79.5%, PPV of 37.1%, NPV of 87.6%, and an accuracy of 74.3% (Table 1).

ROC curve analyses for PIVKA-II showed that the AUC was 0.515; Youden index J was 0.089; the cutoff was 19 mAU/mL with a sensitivity of 85.3% and specificity of 23.6% (Fig. 6C). With the application of an institutional cutoff of 40 mAU/mL, high PIVKA-II was observed in 59 of 109 (54.1%) in the HCC group, and 210 of 381 (55.1%) in the non-HCC group (p=0.855), showing a sensitivity of 54.1%, specificity of 44.9%, PPV of 21.9%, NPV of 77.4%, and an accuracy of 46.9% (Table 1).

Profiles in Patients with Other Diseases

The mean and median AFP level were 30.7±84.0 ng/mL and 9.9 ng/mL respectively in the HCC group, and 45.4±216.8 ng/mL and 3.1 ng/mL respectively in the non-HCC group (p<0.001; Fig. 7A). The mean and median PIVKA-II level were 279.2±855.6 mAU/mL and 30 mAU/mL respectively in the HCC group, and 473.4±2,004.8 ng/mL and 25 mAU/mL respectively in the non-HCC group (p=0.190; Fig. 7B).

Figure 7. Comparison of alpha-fetoprotein (AFP) (A) and protein induced by vitamin K absence or antagonist-II (PIVKA-II) (B) levels according to the presence and absence of hepatocellular carcinoma (HCC) in patients with other liver diseases. Bars indicate 25-75 percentiles. (C) Receiver operating characteristic curve analysis of AFP and PIVKA-II for diagnosis of HCC.

ROC curve analyses for AFP showed that the AUC was 0.696; Youden index J was 0.359; the cutoff was 5.0 ng/mL with a sensitivity of 70.5% and specificity of 65.4% (Fig. 7C). With the application of an institutional cutoff of 7.5 ng/mL, high AFP was observed in 34 of 61 (55.7%) in the HCC group and 86 of 306 (28.1%) in the non-HCC group (p<0.001), showing a sensitivity of 55.7%, specificity of 71.9%, PPV of 28.3%, NPV of 89.1% and an accuracy of 69.2% (Table 1).

ROC curve analyses for PIVKA-II showed that the AUC was 0.515; Youden index J was 0.089; the cutoff was 19 mAU/mL with a sensitivity of 70.5% and specificity of 38.4% (Fig. 7C). With the application of an institutional cutoff of 40 mAU/mL, high PIVKA-II was observed in 25 of 61 (41.0%) in the HCC group, and 120 of 306 (39.2%) in the non-HCC group (p=0.797), showing a sensitivity of 41.0%, specificity of 60.8%, PPV of 17.2%, NPV of 83.8%, and an accuracy of 57.5% (Table 1).

HCC Prediction with the Combination of AFP and PIVKA-II

The combination of institutional cutoffs of 7.5 ng/mL in AFP or 40 mAU/mL in PIVKA-II resulted in a sensitivity of 65.7%, specificity of 45.9%, PPV of 56.4%, NPV of 55.7%, and an accuracy of 56.1% (Table 1). The combination of these two biomarkers did not increase the diagnostic predictability of HCC, compared with a single biomarker.

The results of the present study revealed that serum AFP and PIVKA-II were variably expressed in patients with HCC and in patients with liver cirrhosis without evidence of HCC.

AFP is a 67-kDa glycoprotein that is produced in early fetal life by the liver and by a variety of tumors including HCC, hepatoblastoma, and non-seminomatous germ-cell tumors of the ovary and testis. Tumor cells synthesize fetal proteins following the de-differentiation of adult hepatocytes [7]. During fetal life, AFP is synthesized at first by the yolk sac, then by the liver. By the end of the first trimester, the fetal liver produces nearly all of the AFP. Although synthesis is reduced markedly shortly after birth, small amounts of AFP continue to be produced during adulthood [8]. Normal concentrations of AFP in adult serum are ≤20 ng/mL. AFP can increase temporarily in cases of liver injury or regeneration, particularly after liver resection, during fulminant viral hepatitis, or chronic viral hepatitis [9]. Patients with chronic hepatitis or cirrhosis and persistently elevated AFP levels are at higher risk of developing HCC [10]. Up to 20% of the HCC patients do not produce AFP [11]. Serum AFP levels increase by 20%–80% in patients with HCC and are strongly related to tumor aggressiveness [12]; its concentrations are correlated with tumor size, microvascular invasion, and poorly differentiated HCC [13].

In 1984, it was found that PIVKA-II was significantly increased in the serum of HCC patients and it could serve as a new serum marker for HCC [14]. Many studies suggested that together, as makers for HCC, PIVKA-II, and AFP may improve HCC diagnosis compared to the use of each biomarker alone [15]. The diagnostic value of PIVKA-II is controversial and it is still under discussion whether there is a correlation between PIVKA-II and AFP and whether PIVKA-II can completely replace or supplement the role of AFP in the HCC diagnosis [15,16].

It is reported that serum AFP (>15 or 20 ng/mL) as a screening test for HCC has a sensitivity of between 39% and 64%, a specificity of between 76% and 91%, and PPV of between 9% and 33% [10,17]. A case-control study of 340 cirrhotic patients showed that AFP levels >20 ng/mL had a sensitivity of 60% and a specificity of 91% to diagnose HCC. At this threshold, 40% of HCC cases would be missed [18].

Although the prognostic value of AFP in LT appears to be already established, the cut-off value to define the level of AFP is still a challenge. There is no clear consensus regarding the AFP level above which a patient should not be a candidate for LT. More than 20 studies have tried to define a cutoff value for pretransplant AFP, above which the prognosis would be too impaired to propose LT [19].

An Italian multicenter study including 388 patients with chronic liver disease reported that the overall ROC AUC values for AFP and PIVKA-II were 0.698 (p<0.001) and 0.780 (p<0.001), respectively. AFP and PIVKA-II AUCs were 0.822 and 0.833, respectively, in chronic HBV patients; 0.648 and 0.732 respectively in chronic HCV patients; 0.640 and 0.806 in non-viral chronic liver disease patients. Diagnostic accuracies of AFP and PIVKA-II were 40.5%–59.8% and 62.7%–73.5% and those of combined markers were 78.2% in chronic HBV patients; 77% in non-viral chronic liver disease patients; and 75% in chronic HCV patients. AFP level correlated with alanine transaminase in HCC patients with chronic HCV and non-viral chronic liver disease, but not in chronic HBV infection. PIVKA-II correlated independently with the size of chronic liver disease etiology tumor and AFP in chronic HBV patients only. The diagnostic performance of AFP and PIVKA-II is significantly influenced by the etiology and activity of the chronic liver disease, thus their combination provides a better diagnostic accuracy [20].

In the present study, the expression patterns of PIVKA-II in ALD patients were unique, and PIVKA-II was observed to have no role in the diagnosis of HCC. A Korean study with 2,528 patients without HCC revealed that ALD and antibiotics usage may be confounding factors when interpreting high serum PIVKA-II levels in patients without HCC. Therefore, serum PIVKA-II levels in patients with ALD should be interpreted with caution [21].

The results of the present study revealed that AFP and PIVKA-II are variably expressed according to the HCC status and background liver disease. It is known that the clinical features and mechanisms of carcinogenesis differ quite markedly depending on the underlying diseases. HBV-related HCC more frequently presents as a massive tumor that is more aggressive than HCV-related HCC [22]. With regard to carcinogenesis, integrated HBV-DNA and HBx proteins have been reported to play a critical role in HBV-related HCC, whereas underlying liver cirrhosis and chronic inflammation of the liver are believed to play a role in HCV-associated oncogenesis [23,24]. The present study revealed that a combination of AFP and PIVKA-II did not increase the diagnostic predictability of HCC, compared with a single biomarker.

This study has a number of limitations. First, it was a retrospective single-center study. Second, the severity of liver diseases and the stage of HCC were not included for analysis. Third, pretransplant HCC treatment was also not taken into account.

In conclusion, the results of the present study indicate that serum AFP and PIVKA-II may be expressed variably according to the status of HCC and background liver disease. Serum PIVKA-II was uniquely highly expressed in liver cirrhosis patients with non-viral etiology. Therefore, the values of HCC tumor markers should be cautiously interpreted in liver transplant candidates.

There was no funding related to this study.

All authors have no conflicts of interest to declare.

Conceptualization: CSP. Data curation: CSP, SHK, SH. Methodology: All. Visualization: SH. Writing - original draft: CSP, YKC, SHK. Writing - review & editing: All.