Performance of protein induced by vitamin K absence or antagonist-II assessed by chemiluminescence enzyme immunoassay for hepatocellular carcinoma detection: a meta-analysis.

Abstract Objectives: In the setting of surveillance for hepatocellular carcinoma (HCC) detection, the use of serum biomarkers in addition to ultrasonography (US) is still a matter of debate. Hence, we performed a meta-analysis to evaluate the diagnostic accuracy of protein induced by vitamin K absence or antagonist-II (PIVKA-II) and alpha-fetoprotein (AFP) alone or in combination for HCC detection in patients at risk of tumor development. Materials and methods: We performed a systematic search in PubMed and Scopus database for original articles published in English from 2011 to 2017, investigating the accuracy of PIVKA-II alone or in combination with AFP (reported as area under the curve [AUC]) for HCC detection among patients at risk of tumor development. Furthermore, we focused on studies in which serum PIVKA-II was assessed by highly sensitive chemiluminescence immunoassay (CLEIA). Results: A total of 11 studies (873 patients with HCC and 1244 patients with advanced liver disease/cirrhosis) were included in the meta-analysis. The weighted summary AUC (sAUC) of PIVKA-II and AFP for the discrimination between patients with HCC and those without was 0.791 (0.746–0.837) and 0.767 (0.732–0.803), respectively. The combination of PIVKA-II + AFP results in a sAUC of 0.859 (0.837–0.882). The performance for HCC detection of PIVKA-II + AFP was significantly superior to each biomarker used alone (ΔsAUC = 0.068, p = .032 and ΔsAUC = 0.092, p < .001, respectively). Conclusion: In clinical practice, the use of PIVKA-II + AFP in addition to US examination may improve the effectiveness of surveillance among patients at risk for HCC development.


Introduction
Despite the availability of direct-acting antiviral agents (DAAs) and potent nucleos(t)ide analogs (NAs) with high genetic barrier to resistance for treatment of patients with chronic hepatitis C and B, respectively, the prognosis of patients with advanced liver disease/cirrhosis responder to antiviral treatment is still characterized by a significant risk of hepatocellular carcinoma (HCC) development [1,2].
Currently, surveillance programs for HCC detection in high risk population are mainly based on abdominal ultrasonography (US) [3,4]. However, the United States has no predictive power on HCC development. Indeed, hepatocarcinogenesis is a gradual process characterized by genetic and molecular alterations progressively accumulating within hepatocytes before the appearance of a neoplastic lesion detectable by imaging [5]. Thus, additional tools allowing early HCC detection or prediction are needed to complement US screening. Among traditional serum markers, alpha-fetoprotein (AFP) is the most used worldwide despite showing suboptimal performance for HCC detection [6]. Conversely, novel classes of biomarkers involved in epigenetic machinery such as microRNAs (miRNAs) showed promising results [7,8]. However, their use in clinical practice may be limited by the absence of a standardized analytical method and by a complex miRNA-messenger RNA interference network reflecting different genetic and epigenetic features, that in turn may significantly alter miRNAs expression [9].
Protein induced by vitamin K absence or antagonist-II (PIVKA-II), also known as des-gamma-carboxy prothrombin, is a biomarker specific for HCC firstly described in 1984 [10]. PIVKA-II is an hypo-carboxylated prothrombin released by the liver in absence of vitamin K or in presence of malignant cells; higher PIVKA-II serum levels are associated to tumor size, microvascular invasion and predict HCC recurrence [11][12][13]. The combination of PIVKA-II with AFP is currently used in Japan for surveillance of patients at risk of HCC development as recommended by the local HCC guidelines [14]. However, results regarding PIVKA-II performance in comparison or in combination with AFP are conflicting and available data are mainly derived from studies involving Asiatic patients [15,16]; results from Western studies are limited by the relatively small sample size. Furthermore, methods for PIVKA-II measurement in serum evolved over time from competitive radioimmunoassay and enzyme immunoassay to fully automated chemiluminescence enzyme immunoassay (CLEIA) [17]. Thus, technical improvement may have affected analytical performance.
Considering the availability of novel CLEIA-based methods for PIVKA-II measurement and the lack of consensus regarding the usefulness of PIVKA-II in the setting of HCC surveillance, we performed a meta-analysis on the diagnostic accuracy of PIVKA-II alone or in combination with AFP for HCC detection among patients at risk of tumor development.

Search strategy
The meta-analysis of observational studies in epidemiology (MOOSE) reporting guidelines for the performance of metaanalyses have been followed [18]. Original research articles published in English on accuracy of PIVKA-II for the discrimination of patients with HCC from those with advanced liver disease/cirrhosis were identified through PubMed (https:// www.ncbi.nlm.nih.gov/pubmed) and Scopus (https://www. scopus.com) database. The search strategy was based on the use of the following terms: 'PIVKA-II' or 'des-gamma-carboxy prothrombin' or 'DCP' and 'HCC' or 'hepatocellular carcinoma'. Since we focused on articles in which PIVKA-II was tested by CLEIA method, only papers published from 1 January 2011, were screened. For both databases the search was performed on 30 August 2017. Patients with HCC were considered the 'case group' whereas patients with advanced chronic liver disease or cirrhosis or with non-neoplastic liver nodules were considered as the 'control group'. No restriction was set for age and sex of the patients.

Study selection
Two authors (GPC and DGR) independently reviewed the titles and the abstracts of the studies retrieved from electronic search and selected those potentially relevant for the meta-analysis. The full-text of selected studies was assessed by three authors (GPC, DGR and MLA) to determine whether the inclusion criteria were satisfied.
Inclusion criteria were: (1) original research articles published in English; (2) studies reporting PIVKA-II diagnostic accuracy for the discrimination between patients with HCC and patients with advanced liver disease/cirrhosis in comparison to AFP and (3) studies investigating PIVKA-II performance assessed by CLEIA method. Duplicates studies and studies lacking of data of interest were excluded. The quality of included studies was assessed by the quality assessment of diagnostic accuracy studies (QUADAS) tool [19]. Accordingly, a maximum of 14 points could be awarded answering questions related to biases, variability and reporting.

Data extraction
From selected papers, the same two authors (GPC and DGR) extracted data regarding authors, Country, year of publication, type of study, number of patients, underlying liver disease etiology, biomarkers performance (area under the curve [AUC] and 95% confidence interval [CI]), sensitivity (Se) and specificity (Sp) at the corresponding cutoff value.

Statistical analysis
The meta-analysis was performed using MedCalcV R software version 15.8.1 (Ostend, Belgium). Chi-square test was performed to evaluate difference of categorical variables between groups. Test for inter-rater agreement (Cohen Kappa statistics) was used to evaluate the agreement between investigators. The weighted summary AUC (sAUC) was calculated including each AUC value and the corresponding standard error (SE) from all included studies. SE was calculated with the following formula: SE ¼ (ln UBln LB)/2 Ã 1.96, where UB and LB were the upper and lower bound of the 95% CI of AUC, respectively.
Forrest plots showing the overall effect and funnel plots for publication bias assessment were constructed. According to the presence of heterogeneity, a fixed or random effects model was preferred. Cohcran's Q and I 2 statistics were used to detect heterogeneity; a p value < .1 and I 2 value > 25% were considered as indicative of heterogeneity, respectively.
To measure funnel plot asymmetry, Egger regression analysis was performed [20]. Accordingly, the standard normal deviate, defined as the natural logarithm of estimate divided by its SE, was regressed against the estimate's precision, defined as the inverse of the SE. The intercept of the regression line and the corresponding p value provided the measure of asymmetry.

Results
A total of 11 studies were included in the meta-analysis (Table 1) [21][22][23][24][25][26][27][28][29][30][31]. The strategy search is depicted in Figure 1. There was no disagreement among authors regarding eligibility of original articles finally included in the meta-analysis (K statistics ¼ 1). Overall, 2117 patients were included: 873 patients with HCC (Case group), 683 patients with cirrhosis and 561 with chronic liver disease (Control group). The underlying liver disease etiology was mainly viral, with a significant higher proportion of patients with chronic hepatitis B virus (HBV) infection (p < .001). For each study included, the performance of PIVKA-II and AFP were reported in Table 2.
The sAUC of PIVKA-II for the discrimination between patients with HCC and those without was 0.791 (0.746-0.837; Figure 2). Since the studies showed heterogeneity (p < .001 of Cohcran's Q and I 2 ¼ 81.8%), a random effects model was applied. The sAUC of AFP for the discrimination between patients with HCC and those without tumor was 0.767 (0.732-0.803; Figure 3). Considering that the studies showed heterogeneity (p ¼ .090 of Cohcran's Q and I 2 ¼ 38.8%), a random effects model was applied. The sAUC of PIVKA-II in combination with AFP for the discrimination between patients with HCC and those without was 0.859 (0.837-0.882; Figure 4). A fixed effects model was applied because the results of the studies showed no heterogeneity ( p ¼ .676 of Cohcran's Q  89.4% a PIVKA-II cut-off. b AFP cut-off. c Only cohort B (high risk population surveillance) was included in meta-analysis. AFP: alpha-fetoprotein; AUC: area under the curve; CI: confidence interval; n.a.: not available; PIVKA-II: protein induced by vitamin K absence or antagonist-II; Se: sensitivity; SE: standard error; Sp: specificity. and I 2 ¼ 0%). No differences were observed between PIVKA-II and AFP diagnostic accuracy (DsAUC ¼ 0.024, p ¼ .524), whereas the combination of PIVKA-II þ AFP was significantly superior to each biomarker used alone (DsAUC ¼ 0.068, p ¼ .032 and DsAUC ¼ 0.092, p < .001, respectively).

Publication bias
Egger test for publication bias showed that the risk of having missed or overlooked studies was minimal for the assessment of sAUC of PIVKA-II and AFP (p ¼ .136 and .037, respectively), whereas no publication bias was observed for the analysis of biomarkers combination (p ¼ .522; Figure 5).

Discussion
The use of serum biomarkers for HCC surveillance is a matter of debate since no unequivocal evidence in improving early HCC detection has been produced [32]. Nevertheless, conventional biomarkers such as AFP and PIVKA-II are used in clinical practice, although not universally recommended by scientific guidelines.
AFP is the widest used biomarker in the setting of surveillance of high risk population albeit showing a large range of Se (40-65%) and Sp (76-96%) values for HCC detection [33]. Furthermore, AFP values are often increased in patients with chronic liver disease or cirrhosis and without HCC, as a consequence of inflammation, necrosis and regeneration [34]. Nonetheless, as suggested by Asian Pacific Association for the Study of the Liver, AFP used in combination with US, may be useful to increase Se without decreasing Sp for tumor recognition [35].
To date, several biomarkers have been proposed as potential alternative or diagnostic complement to AFP [16,[36][37][38][39]. Amongst these, PIVKA-II has been extensively investigated either alone or in combination with AFP, but results  concerning performance for HCC detection are conflicting [6,40]. In the present meta-analysis including 11 studies, we found no difference between AFP and PIVKA-II diagnostic accuracy for the discrimination between patients with HCC and those without, but the combination of both biomarkers led to a significant improvement in the performance of HCC detection. Consistently, several studies focused either on the combination of biomarkers or on the development of scores that include different classes of biomarkers and even demographical or clinical characteristics [29,[41][42][43][44][45], in order to improve reliability and performance for HCC detection. As a matter of fact, these strategies showed promising results. Furthermore, some of these scores seem able to accurately predict HCC development among high risk patients [29,46].
Another major issue is represented by the method used for biomarker assessment. Most of novel biomarker proposed for HCC detection have been assessed by non-standardized methods, not allowing to reliably reproduce or compare results from different studies and thus, still far from any potential use in clinical practice [47]. For this reason, only biomarkers evaluated by highly sensitive standardized methods may be recommended.
The results of this meta-analysis may be limited by lack of pathological characterization of HCC cases. Since data regarding HCC classification (i.e., staging according to Barcelona Clinic Liver Cancer system) or nodules features (such as number, size and vascular invasion) were absent or not limited to very early/early tumor stages, we could not assess the performance of biomarkers for early HCC detection. Another potential limitation is represented by the presence of noncirrhotic patients in control group, as cirrhosis is considered the principal risk factor for HCC development. However, we included studies that enrolled as controls, only patients with advanced liver diseases/cirrhosis in order to obtain a control group representative of the real population under surveillance for the risk of HCC development.
In conclusion, the results of this meta-analysis highlight the added value of PIVKA-II and AFP combination for HCC detection rather than a single biomarker used alone. In clinical practice, the use of this combination in addition to US examination may be considered to improve the effectiveness of surveillance of patients at risk for HCC development. Nonetheless, prospective multicenter studies including a large cohort of patients with advanced liver disease/cirrhosis are needed to evaluate PIVKA-II þ AFP performance for tumor  prediction, thus allowing the identification of patients at higher risk of HCC development.