Primary hepatocellular carcinoma (HCC) is one of the most malignant tumors and is the most common cause of cancer-related death worldwide [
European Association for the Study of the Liver (EASL) recommends patients with liver diseases to do liver ultrasound and examine serum α-fetoprotein (AFP) level every six months [
Glypican-3 (GPC3) belongs to the glypican family of heparan-sulfate proteoglycans [
We performed a comprehensive search on the peer-reviewed scientific literatures that are written in English and were published before May 20, 2014, in NCBI PubMed or EMBASE. The following search terms were used: (1) GPC3: glypican-3 and GPC3; and (2) HCC: HCC, hepatocellular carcinoma, liver cancer, liver cell carcinoma, and hepatic cell carcinoma. No restrictions on study design, year of publication, or publication type were set during initial database search. To avoid exclusion of relevant studies, we did not use keywords or indexing terms for diagnostic test accuracy. After filtering the studies based on the criteria listed in the next paragraph, we manually searched the reference lists of selected articles to identify more relevant publications.
The inclusion criteria for articles were as follows: (1) studies investigated the diagnostic accuracy of serum GPC3 for HCC; (2) studies have reported calculable data on sample sizes of HCC and non-HCC patients, true positive (TP), true negative (TN), false positive (FP), and false negative (FN) values; and (3) article is written in English. The exclusion criteria were as follows: (1) studies conducted on animals; (2) studies that evaluated mRNA expression or DNA polymorphisms of GPC3 and did not provide the sensitivity or specificity of using GPC3 as a HCC marker; (3) letters, editorials, expert opinions, and reviews without original clinical data; (4) case reports and studies lacking control groups; and (5) duplicate reports.
Initial screening for potentially eligible studies was carried out by author Sheng-Li Yang based on the titles and abstracts of articles. Two authors Sheng-Li Yang and Xiefan Fang independently reviewed and included eligible studies based on the criteria described above. Disagreements were resolved by discussion or consulting with author Zao-Zao Huang. After all the eligible studies were identified, the following characteristics were retrieved from each study: authors, geographic distribution of patients, study design, number of patients, reference test, methods of measurement, cutoff values, and raw data including TP, FP, TN, and FN results.
Authors Sheng-Li Yang and Xiefan Fang independently assessed qualities of the eligible studies using the recommended checklist of Quality Assessment of studies of Diagnostic Accuracy included in Systematic reviews (QUADAS, Cochrane Collaboration). Each of the eleven items in the QUADAS checklist was scored as “yes,” “no,” or “unclear” [
For each study, we calculated sensitivity, specificity, positive likelihood ratios (LR+), and negative likelihood ratios (LR−), as well as their corresponding 95% confidence intervals (95% CIs). The data were visualized as forest plots and receiver operating characteristic curves (ROC). Heterogeneity of the retrieved data from eligible studies was evaluated by using the Q statistics, with a significance level at
Because substantial heterogeneity existed in the included studies, we performed subgroup meta-analysis by dividing studies into one group that found that GPC3 has diagnostic value in HCC detection and another group that claimed that GPC3 has no diagnostic value for HCC. The Spearman approach was applied to test if the heterogeneity can be explained by a threshold effect.
A total of 823 potentially relevant articles were identified by searches in NCBI PubMed and Embase. After reviewing their titles and abstracts, 523 articles, including duplicate studies, case reports, reviews, and comments, were excluded. After reviewing the full texts, 256, 20, and 2 studies were excluded due to irrelevant study design, insufficient data to estimate sensitivity or specificity, and publishing overlapping data, respectively. In the references of the retrieved studies, no additional articles met our inclusion criteria. Finally, twenty-two studies were included for meta-analysis [
Characteristic and methodology assessment of the included studies.
First author, year, country | Characteristics of HCC | Characteristics of controls | GPC3 | ||||
---|---|---|---|---|---|---|---|
Assay type | Cut-off value | Antibody for detection | HCC value (ng/mL) | Controls value (ng/mL) | |||
Yu [ |
NA | LC or hepatitis patients; healthy individuals | Chemiluminescent immunoassay | 30 ng/mL | GPC3 8G6 mcAb and 7D11 mcAb (Millipore Corporation) | 108.67 ± 230.04 | 3.99 ± 7.68 |
|
|||||||
Lee [ |
62.5% were HBV associated | CLD (HCV) patients; 50% had liver cirrhosis | ELISA | 73 ng/mL | ELISA kit (Wuhan Cusabio Biotech) | 75.8 ± 117.5 | 66.4 ± 33.2 |
|
|||||||
Badr [ |
NA | LC (HCV) patients | ELISA | 240 ng/mL | ELISA kit (Wuhan Uscn) | 551.47 ± 185.25 | 98.23 ± 73.54 |
|
|||||||
Li [ |
HBV associated HCC | Healthy individuals | ELISA | NA | ELISA kit (usabio Biotech) | 12.63 ± 2.93 for patients with AFP <400 |
1.92 ± 0.95 |
|
|||||||
Chen [ |
NA | LC or hepatitis patients; healthy individuals | ELISA | 25.25 ng/mL | 7C8 and GP9 mcAb (self-made) | 99.94 ± 267.2 | 19.44 ± 50.88 for LC patients; |
|
|||||||
Abdelgawad [ |
67.5% were HCV associated; 17.5% were HBV associated | LC patients; healthy individuals | ELISA | 4.9 ng/mL | ELISA kit (Wuhan Uscn) | 7.7 | 3.24 |
|
|||||||
Gomaa [ |
12.9% were HBV associated; 87.1% were HCV associated | LC patients with HCV/HBV | ELISA | 5.41 ng/mL | ELISA kit |
8.13 ± 3.25 | 3.14 ± 1.16 |
|
|||||||
Wang [ |
HBV associated | LC patients with HBV | ELISA | NA | ELISA kit |
NA | NA |
|
|||||||
Qiao [ |
76.2% were HBV associated; 7.9% were HCV associated | LC (HBV/HCV) or hepatitis patients; healthy individuals | ELISA | 26.8 ng/mL | ELISA kit |
29.29 ± 17.34 | 12.09 ± 9.69 for LC patients; 9.98 ± 9.60 for chronic hepatitis patients; 5.93 ± 5.46 for healthy controls |
|
|||||||
Abd El Moety [ |
HCV associated | LC or hepatitis patients; healthy individuals | ELISA | 2.0 ng/mL | ELISA kit |
34.63 ± 23.8 | NA |
|
|||||||
Zhang [ |
NA | LC or hepatitis patients with HCV/HBV; healthy individuals | Immunoassay | 3.10 ng/mL | Self-made | 116.8 ± 98.6 | 24.60 ± 24.01 for LC patients; 6.73 ± 12.2 for hepatitis B patients; 13.67 ± 15.68 for hepatitis C patients; 0.86 ± 1.12 |
|
|||||||
Youssef [ |
HCV and HBV associated | LC patients with HBV/HCV; |
ELISA | 4.6 ng/mL | ELISA kit |
NA | NA |
|
|||||||
Liu [ |
84% were HBV associated; 16% were HCV associated | LC patients with HBV/HCV | ELISA | 300 ng/L | ELISA kit |
NA | NA |
|
|||||||
Tangkijvanich [ |
59% were HBV associated; 11% were HCV associated | LC or hepatitis patients with HBV/HCV | ELISA | NA | Self-made | 46.3 (0–7826.6)■ | 0 (0–43.6)■ |
|
|||||||
Beale [ |
60% had ALD and 40% patients had NAFLD | LC patients with ALD/NAFLD | ELISA | NA | ELISA kit |
161.41 ± 422.33 | 125.41 ± 281.05 |
|
|||||||
Yoshitaka Hippo [ |
NA | LC patients | ELISA | 2.0 ng/mL | Self-made | 4.84 ± 8.91 | 1.09 ± 0.74 |
|
|||||||
Nakatsura [ |
12.1% were HBV associated; 40.9% were HCV associated | LC patients with HBV/HCV/PBC/AIH | ELISA | 10 U/mL | Self-made | NA | NA |
|
|||||||
Capurro [ |
44.1% were HBV associated; 27.2% were HCV associated; 14.7% had ALD | LC or hepatitis patients with HBV/HCV | ELISA | 117 ng/mL | Self-made |
NA | NA |
|
|||||||
Wang [ |
84.5% were HBV associated; 9.5% were HBV associated; 3.5% were alcoholic cirrhosis associated | LC or hepatitis (HBV) patients; healthy individuals | ELISA | NA* | ELISA kit (usabio Biotech) | 4.55 ± 3.5 | 9.03 ± 4.1 for LC patients; 4.56 ± 3.2 for HBV chronic hepatitis patients; 6.71 ± 1.8 for healthy controls |
|
|||||||
Nault [ |
With alcoholic cirrhotic, 46.4% were early stage | Alcoholic cirrhotic patients | ELISA | NA* | ELISA kit (Wuhan Cusabio Biotech) | 1.4 ± 0.5 (early stage) |
2.5 ± 1.4 |
|
|||||||
Özkan [ |
52% were HBV associated; 23% were HCV associated | LC patients with HBV/HCV/HDV/AIH/Wilson disease/Cryptogenic liver disease; healthy individuals | ELISA | NA* | ELISA kit |
5.13 ± 22.7 | 5.51 ± 59.2 |
|
|||||||
Yasuda [ |
16% were HBV associated; 77.5% were HCV associated | CLD patients with HBV/HCV | ELISA | NA* | ELISA kit |
924.8 (495.2, 1335.6)▲ | 1161.6 (762.0, 1784.0)▲ |
NA: data are not available; HBV: hepatitis B virus; HCV: hepatitis C virus; LC: liver cirrhosis; CLD: chronic liver disease; ALD: alcoholic liver disease; ALC: alcoholic liver cirrhosis; NAFLD: nonalcoholic fatty liver diseases; PBC: primary biliary cirrhosis; AIH: autoimmune hepatitis; ▲mean median (25% and 75% quartiles); ■mean median (ranges); *cutoff values were not provided because these studies found serum GPC3 has no correlation with HCC.
Study selection process.
The results of QUADAS quality assessment of the included studies are shown in Figure
Summary of methodological quality of included studies on the basis of review authors’ judgments on the 11 items of QUADAS checklist for each study.
Among the twenty-two studies, eighteen of them have demonstrated that serum GPC3 level is higher in HCC patients than that in control subjects, which include healthy individuals and patients with hepatitis or liver cirrhosis [
Patients enrolled in the selected studies used for meta-analysis.
Author (ref.) | Case | Control | TP | FP | FN | TN |
---|---|---|---|---|---|---|
Yu et al. [ |
192 | 101 | 104 | 1 | 88 | 100 |
Lee et al. [ |
120 | 40 | 65 | 14 | 55 | 26 |
Badr et al. [ |
30 | 30 | 30 | 2 | 0 | 28 |
Li et al. [ |
605 | 25 | 330 | 5 | 275 | 20 |
Chen et al. [ |
155 | 440 | 62 | 27 | 93 | 413 |
Abdelgawad et al. [ |
40 | 20 | 38 | 1 | 2 | 19 |
Gomaa et al. [ |
31 | 30 | 28 | 1 | 3 | 29 |
Wang et al. [ |
78 | 97 | 28 | 44 | 50 | 53 |
Qiao et al. [ |
101 | 88 | 52 | 6 | 49 | 82 |
Abd El Moety et al. [ |
10 | 40 | 10 | 24 | 0 | 16 |
Zhang et al. [ |
36 | 93 | 33 | 0 | 3 | 56 |
Youssef et al. [ |
40 | 40 | 33 | 2 | 7 | 38 |
Liu et al. [ |
75 | 32 | 35 | 2 | 40 | 30 |
Tangkijvanich et al. [ |
100 | 100 | 53 | 1 | 47 | 99 |
Beale et al. [ |
50 | 41 | 34 | 22 | 16 | 19 |
Hippo et al. [ |
69 | 38 | 35 | 4 | 34 | 34 |
Nakatsura et al. [ |
40 | 50 | 16 | 0 | 24 | 50 |
Capurro et al. [ |
34 | 91 | 18 | 1 | 16 | 90 |
Wang et al. [ |
84 | 173 | NA | NA | NA | NA |
Nault et al. [ |
125 | 170 | NA | NA | NA | NA |
Özkan et al. [ |
75 | 55 | NA | NA | NA | NA |
Yasuda et al. [ |
200 | 200 | NA | NA | NA | NA |
TP: true positive; FP: false positive; TN: true negative; FN: false negative; NA: data are not available.
Forest plots of sensitivity and specificity of using GPC3 as a diagnostic marker for hepatocellular cancer (HCC) in the eighteen studies included for meta-analysis [
Summary receiver operating characteristic curves (SROC) from the hierarchical summary receiver operating characteristic model generated from the eighteen studies that found that GPC3 is a diagnostic marker for HCC [
We used Deeks’ funnel plot asymmetry test to evaluate publication bias among the included studies. The slope coefficient of the regression line had a
Linear regression test of funnel plot asymmetry. The statistically nonsignificant
HCC is one of the most lethal malignancies with a survival rate less than 10% and its incidence is increasing worldwide [
In this meta-analysis, we identified twenty-two studies that have investigated the diagnostic accuracy of serum GPC3 for HCC. Eighteen of them demonstrated that GPC3 is an ideal HCC diagnostic marker with pooled sensitivity, specificity, LR+, and LR− of 69%, 94%, 10.50, and 0.34, respectively [
We found that the infection status of HBV or HCV in HCC and control subjects were different in the included studies. In the studies from Egypt and Japan, more than 60% of the enrolled patients with HCC were infected with HCV, and approximately 15–20% of patients were infected with HBV [
The assay reagents used to measure serum GPC3 levels differed among the included studies. Some studies used self-made antibodies. For example, Hippo et al. used an antibody that binds to the N-terminal portion of GPC3 cleaved at Arg358 (amino acids 25–358) [
We noticed that one common experimental design of the four conflicting studies is that they enrolled liver cirrhosis patients as control subjects, even though they had various complications, including alcoholic cirrhosis and HBV- or HCV-associated cirrhosis [
Considering that the included studies have substantial heterogeneity and part of them held the opposite views, a well-designed prospective study with larger cohorts should be performed to rigorously evaluate the diagnostic accuracy of GPC3 and determine if it has a better diagnostic value compared with other known HCC serum biomarkers. In addition, experimental design should be improved in the following areas: (1) double-blind studies should be designed to avoid bias; (2) the cohorts of healthy individuals, hepatitis patients, and liver cirrhosis patients should be compared as separate groups; (3) the study could use two or more different GPC3 antibodies to measure GPC3 level; (4) it is important to examine the stability of GPC3 during long-term storage. Half of the included studies measured GPC3 level in frozen serum, but it is not sure if GPC3 would be degraded after long-term storage. If the stability of serum GPC3 in long-term storage is not good, the diagnostic performance of serum GPC3 may be greatly affected.
To our knowledge, several studies have performed comprehensive reviews on using serum GPC3 as a diagnostic indicator for HCC [
In summary, our meta-analysis indicates that serum GPC3 level is elevated in HCC patients compared with healthy individuals. But whether GPC3 is an index to differentially diagnose HCC and liver cirrhosis is still uncertain.
The authors declare that there is no conflict of interests regarding the publication of this paper.
Sheng-Li Yang and Xiefan Fang contributed equally to this work.
The authors thank Professor Yi Sun (School of Public Health, Tongji Medical College, Huazhong University of Science and Technology) for her valuable suggestions. This study was supported by the National Natural Science Foundation of China (no. 81272421).