Alpha-Fetoprotein, Protein Induced by Vitamin K Absence or Antagonist II and Glypican-3 for the Detection and Prediction of Hepatocellular Carcinoma in Patients with Cirrhosis of Viral Etiology

Simple Summary Circulating biomarkers for the early detection and prediction of hepatocellular carcinoma development are an unmet need. In the present study, we observed that serum values of three biomarkers (namely AFP, PIVKA-II and GPC-3) were significantly different between patients with cirrhosis and those with hepatocellular carcinoma; the best accuracy for the detection of tumors was achieved by a combination of AFP + PIVKA-II. However, PIVKA-II resulted as the only biomarker able to identify patients with cirrhosis at increased risk of hepatocellular carcinoma development. The measurement of PIVKA-II in patients with cirrhosis at risk of tumor development may be useful to tailor personalized surveillance strategies and thus to improve patients’ survival. Abstract International guidelines recommend the use of ultrasound as a surveillance tool for hepatocellular carcinoma (HCC) in patients with cirrhosis, while the role of serum biomarkers is still debated. We investigated serum alpha-fetoprotein (AFP), protein induced by vitamin K absence or antagonist II (PIVKA-II) and glypican-3 (GPC-3) diagnostic accuracy for HCC detection and prediction in patients with liver cirrhosis of viral etiology under surveillance. A total of 349 patients (200 cirrhosis and 149 HCC) were enrolled. The 200 patients with cirrhosis consisted of 114 patients still HCC-free after 36 months of follow-up and 86 patients that developed HCC after 13.8 (11.0–19.8) months. AFP, PIVKA-II and GPC-3 were measured in serum samples collected at tumor diagnosis in the 149 patients with HCC, and at the beginning of follow-up in the 200 patients with cirrhosis. The higher performance for HCC detection was observed for PIVKA-II (area under the curve (AUC) = 0.790), followed by AFP (AUC = 0.737) and GPC-3 (AUC = 0.637); the combination of AFP + PIVKA-II improved the diagnostic accuracy to AUC = 0.822. Serum PIVKA-II values, but not AFP and GPC-3, were significantly higher in the 86 cirrhotics that developed HCC compared with the 114 cirrhotics still HCC-free after 36 months of follow-up (p = 0.020). PIVKA-II ≥ 55 mAU/mL allowed to identify patients with cirrhosis at higher risk of HCC development (Log-rank test, p < 0.001; adjusted Hazard Ratio = 1.99, p = 0.001). In conclusion, the measurement of PIVKA-II in patients with cirrhosis may be useful to tailor personalized surveillance strategies.


Diagnostic Accuracy of AFP, PIVKA-II and GPC-3 for the Detection of HCC
The diagnostic accuracy of AFP, PIVKA-II and GPC-3 for the discrimination between patients with cirrhosis and those with HCC was assessed by receiver operating characteristic (ROC) curve analysis. The values of area under the curve (AUC), sensitivity (Se), specificity (Sp), positive likelihood ratio (+LR) and negative likelihood ratio (−LR) are reported in Table 3. PIVKA-II showed the higher performance with AUC = 0.790, followed by AFP (AUC = 0.737) and by GPC-3 (AUC = 0.637). The best performance derived from the combination of AFP + PIVKA-II (AUC = 0.822) ( Figure  6A). No further improvement was observed following the addition of GPC-3 to AFP + PIVKA-II.

Diagnostic Accuracy of AFP, PIVKA-II and GPC-3 for the Detection of HCC
The diagnostic accuracy of AFP, PIVKA-II and GPC-3 for the discrimination between patients with cirrhosis and those with HCC was assessed by receiver operating characteristic (ROC) curve analysis. The values of area under the curve (AUC), sensitivity (Se), specificity (Sp), positive likelihood ratio (+LR) and negative likelihood ratio (−LR) are reported in Table 3. PIVKA-II showed the higher performance with AUC = 0.790, followed by AFP (AUC = 0.737) and by GPC-3 (AUC = 0.637). The best performance derived from the combination of AFP + PIVKA-II (AUC = 0.822) ( Figure 6A). No further improvement was observed following the addition of GPC-3 to AFP + PIVKA-II.    Afterwards, we assessed the diagnostic accuracy of the biomarkers for the discrimination between patients with cirrhosis and patients with early HCC (BCLC 0 and A, n = 115). Overall, we observed only a moderate reduction of performance (Table S1); the combination of AFP + PIVKA-II maintained a good accuracy, showing AUC = 0.802 ( Figure 6B).
Since the two groups of patients (i.e., patients with cirrhosis and patients with HCC) showed Afterwards, we assessed the diagnostic accuracy of the biomarkers for the discrimination between patients with cirrhosis and patients with early HCC (BCLC 0 and A, n = 115). Overall, we observed only a moderate reduction of performance (Table S1); the combination of AFP + PIVKA-II maintained a good accuracy, showing AUC = 0.802 ( Figure 6B).
Since the two groups of patients (i.e., patients with cirrhosis and patients with HCC) showed significant differences regarding demographic and biochemical features, we performed a multivariate logistic regression analysis to test whether the observed differences may impact on the performance of AFP, PIVKA-II and GPC-3 (Table 4). We observed that AFP and PIVKA-II, but not GPC-3, resulted significantly and independently associated with the diagnosis of HCC.

Prediction of HCC Development in Patients with Cirrhosis under Surveillance
The group of 200 patients with cirrhosis, HCC-free at the time of blood sampling and biomarkers testing, consisted of 114 patients still HCC-free after 36 months of follow-up and 86 patients that developed HCC after a median follow-up of 13.8 (11.0-19.8) months. The characteristics of the two subgroups of patients are reported in Table 5. Patients with cirrhosis that developed HCC during the follow-up were older than patients that did not develop the tumor (65, range 35-82 years vs. 57, range 33-82 years; p < 0.001) and there was a trend towards a higher male prevalence in the former group compared to the latter (70/114 vs. 64/86; p = 0.068). Regarding biochemical features, patients that developed HCC had a lower platelet count compared to patients still HCC-free after 36 months of follow-up (p < 0.001). At baseline, no differences in AFP and GPC-3 serum levels were observed between patients that developed HCC or not (p = 0.851 and p = 0.844, respectively). Conversely, PIVKA-II serum values were significantly higher in patients that developed HCC compared to those still HCC-free after 36 months of follow-up ( Figure S1).    By a Youden J statistic, we identified a PIVKA-II cut-off value of 55 mAU/mL for data dichotomization and subsequent analysis. Kaplan-Meier analysis showed different survival curves ( Figure 7); among the 128 patients with baseline serum PIVKA-II < 55 mAU/mL, 45 (35.2%) developed HCC, while among the 72 patients with baseline serum PIVKA-II ≥ 55 mAU/mL, 41 (56.9%) developed the tumor during the 36 months follow-up (p = 0.001). At multivariate Cox proportional hazards regression analysis corrected for age and platelet count, PIVKA-II ≥ 55 mAU/mL resulted in a significant independent factor associated to HCC development (hazard ratio (HR) = 1.99, 95% CI 1.25-3.19, p = 0.004) ( Table 6).

Discussion
The early detection of HCC, or even its prediction, represents the major goal in order to improve patients' survival; thus, the identification of biomarkers able to reflect molecular alterations preceding tumor development is an unmet need. In the present study, we observed that serum values of AFP, PIVKA-II and GPC-3 were significantly different between patients with cirrhosis and those At multivariate Cox proportional hazards regression analysis corrected for age and platelet count, PIVKA-II ≥ 55 mAU/mL resulted in a significant independent factor associated to HCC development (hazard ratio (HR) = 1.99, 95% CI 1.25-3.19, p = 0.004) ( Table 6).

Discussion
The early detection of HCC, or even its prediction, represents the major goal in order to improve patients' survival; thus, the identification of biomarkers able to reflect molecular alterations preceding tumor development is an unmet need. In the present study, we observed that serum values of AFP, PIVKA-II and GPC-3 were significantly different between patients with cirrhosis and those with HCC; however, the best accuracy for the detection of tumors was achieved by the combination of AFP + PIVKA-II. Furthermore, PIVKA-II resulted to be the only biomarker able to identify patients with cirrhosis at increased risk of HCC development.
To date, several candidate biomarkers have been investigated in order to improve the detection of HCC in patients with cirrhosis under surveillance [16,17]. Among novel biomarkers, non-coding circulating RNAs such as microRNAs, showed interesting results [18]; however, their main limitation is represented by the absence of a standardized method for their quantitation in body fluids [19]. On the other hand, traditional biomarkers such as AFP have been intensely debated due to their suboptimal accuracy. Nonetheless, recent technical advancements in the analytical methods led to an improvement of their performance [20]. In addition, different studies suggested that the combination of different classes of biomarkers and the use of scores based on different clinical or demographic features further improved early HCC detection rates [21][22][23][24][25]. In our study, in agreement with previous findings [10,[26][27][28], the combined use of AFP + PIVKA-II led to a good diagnostic accuracy for the detection of HCC (AUC = 0.822); even including in the analysis only patients with early HCC, the combined performance of AFP + PIVKA-II remained satisfactory (AUC = 0.802). On the other hand, GPC-3 showed only a moderate diagnostic accuracy. Previous reports showed contradictory results concerning the accuracy of GPC-3 for the detection of HCC [17,[29][30][31]; as a matter of fact, the control population included in the studies (i.e., patients with cirrhosis or chronic liver disease or healthy subjects), the overall clinical and biochemical characteristics of the patients enrolled, the different assays used for the measurement of GPC-3 may have accounted for these conflicting findings.
Although the combined used of AFP + PIVKA-II showed a good performance for the detection of HCC, the added value of a non-invasive biomarker is represented by the ability to predict tumor development before imaging discovery. Recently, Loglio and colleagues investigated the performance of PIVKA-II alone or in combination with AFP in 212 nucleos(t)ide analogues-treated patients with HBV-related cirrhosis (64 HCC and 148 HCC-free controls) for the early detection of HCC during surveillance [32]. Interestingly, serum PIVKA-II values raised above the identified cut-off (82 mAU/mL) in 53% of the patients already 6-18 months before HCC detection, showing an accuracy of 86% (Se = 54% and Sp = 100%); combining PIVKA-II with AFP, the accuracy further improved to 90% (Se = 67% and Sp = 100%). Similarly, a real-world study carried out in China showed that the measurement of serum PIVKA-II in patients with HBV-related cirrhosis allowed to stratify the patients according to the cumulative incidence of HCC in a two-years follow up [33]; HCC developed in the 82.0% of patients with elevated PIVKA-II values compared to the 54.1% of patients with low PIVKA-II (p < 0.001). Another recent study by Ricco and colleagues investigated the time-related variability of AFP and PIVKA-II in serum samples of 418 patients with cirrhosis (124 HCC and 294 HCC-free controls) predominantly of viral etiology undergoing standard HCC surveillance by US [34]. It has been observed that both AFP and PIVKA-II increased over time in patients that developed HCC, but among HCC-free controls the time-related changes of PIVKA-II were more stable than AFP; a 5% increase of Log 10 PIVKA-II serum values resulted significantly for the prediction of HCC development. Consistently, we found that serum PIVKA-II levels were significantly different between patients with cirrhosis that developed HCC during 36 months follow-up from those still HCC-free at the end of the observation period. In our series, PIVKA-II, with a cut-off value set at 55 mAU/mL, was able to early identify patients at higher risk of HCC development (HR = 1.99); indeed, the incidence of HCC during the observation period was almost double in patients with baseline serum PIVKA-II ≥ 55 mAU/mL. However, the issue concerning the optimal cut-off for PIVKA-II remains unsolved and the heterogeneity among the different studies is not negligible. The most likely explanation for this discrepancy could lie in the diverse ethnicity, clinical characteristics (such as cirrhosis prevalence, liver disease activity, HCC stage) and any underlying liver disease etiology of the patients enrolled in the different studies; last but not least, the analytical method used for biomarker determination may have impacted on the obtained results.

Study Limitations
The major limitations of our study are represented by the retrospective design and the lack of a cost-effectiveness analysis. Consequently, we were not able to evaluate the benefits deriving from the combination of US and serum biomarkers for HCC surveillance in terms of improved accuracy, better survival and increased quality-adjusted life years. Furthermore, there were several differences between cases and controls concerning demographic and biochemical features. Despite it being well known that older age, male gender and disease activity are important risks factors associated to a higher likelihood of developing HCC, our analysis was adjusted for these potential confounders. Thus, we believe that the results are reasonably robust and provide further evidences concerning the usefulness of PIVKA-II in the setting of HCC surveillance. Further prospective studies including consecutive patients with cirrhosis under surveillance for the risk of HCC development are needed to concretely prove the usefulness of PIVKA-II measurement in the real-life clinical setting.

Patients
This retrospective study included patients with cirrhosis of viral etiology with and without a new diagnosis of HCC, recruited at the outpatient clinic of the Unit of Gastroenterology of Città della Salute e della Scienza di Torino-Molinette Hospital, Turin, Italy, between November 2012 and January 2018.
For all patients, the inclusion criteria were age ≥18 years, serum hepatitis B surface antigen-positivity or anti-HCV-positivity, diagnosis of cirrhosis, availability of a serum sample at the time of HCC diagnosis and signed written informed consent. In addition, for patients with cirrhosis under surveillance, a minimum of 3-year US follow-up after the collection of serum sample was required. Patients in anticoagulant therapy were excluded.
The presence of cirrhosis was determined by liver elastography (FibroScan ® , Echosens™, Paris, France) or by hepatic ultrasound features and endoscopic signs of portal hypertension [35,36]. The diagnosis of HCC was achieved by histological examination or by contrast-enhanced imaging methods showing the radiological hallmark of HCC (i.e., the combination of hypervascularity in late arterial phase and washout on portal venous and/or delayed phases), following international guidelines [5]. HCC was classified according to the BCLC staging system (0 = very early; A = early; B = intermediate; C = advanced; D = end-stage) [5].
Serum samples were collected in polypropylene 2 mL tubes labelled with the study participant identification code and stored at −80 • C until analysis. Study procedures were compliant with the principles of the Declaration of Helsinki. All patients gave their written informed consent and the study was approved by the Institutional Ethics Committee (CEI-452).

Measurement of Serum AFP, PIVKA-II and GPC-3
Serum levels of AFP and PIVKA-II were determined on the fully automated chemiluminescent enzyme immunoassay (CLEIA) system Lumipulse ® G600 II (Fujirebio Inc., Tokyo, Japan) using Lumipulse ® G AFP-N and Lumipulse ® G PIVKA-II reaction cartridges according to manufacturer's instructions. AFP concentrations were given in ng/mL, while PIVKA-II values were reported in mAU/mL [37].
Detection limit of AFP and PIVKA-II assays were 0.075 ng/mL and 1.37 mAU/mL, respectively.

Statistical Analysis
Continuous variables were expressed as median and interquartile ranges (IQR). D'Agostino-Pearson test was used to test data normality. Mann-Whitney test and Fisher's exact test were used to compare continuous and categorical variables, respectively. Kruskal-Wallis test was used to compare continuous variables among more than two independent groups. Correlation between continuous variables was assessed by Spearman's rank correlation coefficient (r s ). To evaluate diagnostic performance of AFP, PIVKA-II and GPC-3 alone or in combination, AUC values, Se, Sp, +LR and −LR were assessed by ROC curves analysis. Multivariate logistic regression analysis was performed to evaluate whether selected variables were significantly and independently associated with HCC (dependent variable). The comparison between survival curves was performed by the Kaplan-Meier method; the difference between the curves was evaluated by Log-rank test. Multivariate Cox proportional-hazards regression was performed to calculate the Hazard Ratios (HR) for HCC development.

Conclusions
In conclusion, PIVKA-II showed a higher diagnostic accuracy than AFP and GCP-3 for the discrimination between patients with or without HCC; furthermore, the combination of AFP + PIVKA-II was superior to PIVKA-II alone for HCC detection. However, the possibility to early identify patients at higher risk of HCC development prior to tumor detection is crucial in order to tailor personalized surveillance strategies and thus to improve survival. The evidences provided by this study suggest that the measurement of serum PIVKA-II may allow to stratify patients with cirrhosis under surveillance according to the likelihood of tumor development and may be useful to select patients that could benefit from a closer monitoring.