Clinical role of combining alpha‐fetoprotein and lens culinaris agglutinin‐reactive fraction of alpha‐fetoprotein for hepatocellular carcinoma: Evidence from literature and an original study

Abstract Background To evaluate the clinical diagnostic efficacy of the combination of alpha‐fetoprotein (AFP) and lens culinaris agglutinin‐reactive fraction of AFP/total AFP (AFP‐L3%) for detecting hepatocellular carcinoma (HCC). Methods A comprehensive and systemic literature search was executed in Web of Science, PubMed, and the Cochrane Library websites. Then, the related articles were reviewed and the quality of included studies was evaluated with the QUADAS tool. Further, serum samples were collected from 49 HCC patients, 52 cirrhosis patients, 47 hepatitis patients, and 48 healthy controls and these samples were tested for AFP and AFP‐L3% levels. Results A total of 16 eligible articles were included in our meta‐analysis. The overall sensitivity (SEN) of AFP + AFP‐L3% was higher than that of AFP or AFP‐L3 alone; the overall specificity (SPE) of AFP + AFP‐L3% was lower than that of AFP or AFP‐L3 alone. In the original study, the related statistics were, respectively, SEN = 0.592 and SPE = 0.918 for AFP; SEN = 0.367 and SPE = 1.000 for AFP‐L3%; and SEN = 0.592 and SPE = 0.918 for the combination. Conclusion The results of meta‐analysis indicate there is a beneficial effect of using the unity of AFP and AFP‐L3% for HCC diagnosing. However, in the original study, just for the results of sensitivity and specificity, there is no significant difference between AFP alone and AFP + AFP‐L3%.


| Data extraction and quality assessment
The first author; year of publication; country of the first author; number of individuals with HCC; individuals with non-HCC (benign liver disease or healthy individuals); study methods; cutoff values; and original data concerning true-positive (TP), false-positive (FP), falsenegative (FN), and true-negative (TN) results were extracted from the eligible studies. We applied the Quality Assessment of studies of Diagnostic Accuracy Included in Systematic reviews (QUADAS) 22,23 to systematically evaluate the quality of the involved studies. A total of 14 items were included, each with response options "yes," "no," or "unclear". A response of "yes" was given a point, while both, "no" and "unclear" scored zero. A QUADAS score ≥9 was considered to indicate that the article was of superior quality.

| Statistical analyses
The statistical analyses were conducted applying the two statistical software programs as below: Stata 12.0 (Stata Corporation) and Meta-Disc software (version 1.4). In the meta-analysis, the I 2 test was used to evaluate the heterogeneity. A probability value of I 2 ≥ 50% and a P value < .1 were regarded as indicative of significant heterogeneity. A random effects model or a fixed effects model was chosen based on the outcomes of the heterogeneity analyses. When the results of I 2 > 50% and P < .05, the random effects model was applied. While a fixed effects model was selected if I 2 was ≤ 50% and P was ≥.05. In this study, four indices, including TP, FP, FN, and TN, were applied to calculate the overall sensitivity, specificity, positive/ negative likelihood ratio (PLR/NLR), diagnostic odds ratio (DOR), 95% confidence interval (95% CI). A summary receiver operating characteristic curve (SROC) and area under the curve (AUC) were graphically and intuitively applied to describe the correlation of sensitivity and specificity. To explore the heterogeneity, the threshold effect and meta-regression analysis were applied. Begg's funnel plot and Egger's liner regression test were used to analyze the publication bias. The AFP concentrations were tested using an electrochemiluminescence immunoassay assay (ECLIA). The serum levels of AFP-L3 were determined using ECLIA after using affinity chromatography assay for AFP-L3 separation. The affinity matrix coupling with lens culinaris agglutinin was fitted on the centrifuge tubes of affinity adsorption, the affinity matrix could combine with AFP-L3 specifically; when the testing samples traversed the centrifuge tubes, AFP-L3 in the samples combined with the affinity matrix and remained in the centrifuge tubes; after eluting, AFP-L3 in the samples was obtained; than using a quantitative analysis on automated platform as AFP to determine the levels of samples, and calculating the AFP-L3/total AFP (AFP-L3%) in final. According to the manufacturer's instructions, the cutoff value was 11ng/ml for AFP and >10% for AFP-L3%.

| Statistic method
In the original study, data analyses were realized using SPSS 20 (SPSS, Inc). There was a comparison of the classified variables performing with chi-square test or Fisher's exact test. It was deemed to statistically significant if there was a 2-tailed P < .05.

| Literature search and features
A total of 422 articles were retrieved from the original database searches; 175 duplicated studies were removed. After the titles and abstracts were screened, 149 articles were excluded because they were reviews, meta-analyses, or unrelated to our study. After reading the remaining 98 full-text articles, 82 were excluded in accordance with the exclusion criteria (10 studies were published in non-English language, 71 studies lacked data to form a 2 × 2 table or lacked key information, and one study had <20 subjects with HCC).
Ultimately, we included 16 set of data in the current meta-analysis  Table 1).
The QUADAS tool was used to identify the quality of the articles, as shown in Table 2. According to the consequences of the methodological and systematic evaluation, the entire included articles were of acceptable quality.

| Meta-analysis
The random effect model was used because all of I 2 values were >50%  (Table 3). These analyses demonstrated that AFP combining with AFP-L3%, rather than either AFP or AFP-L3% alone, has better diagnostic F I G U R E 1 Flowchart of the study selection strategy sensitivity for HCC. The AFP-L3% showed a more superior diagnostic efficiency than AFP in this meta-analysis.

| Heterogeneity and sensitivity analyses
The threshold effect was determined to identify the underlying ori- To determine whether the individual study affected the overall results, the sensitivity analysis was conducted. We found that the individual study had little impact on the final results, indicating that our analyses were stable and reliable.

| Publication bias
To appraise the publication bias of the involved studies, Begg's funnel plot and Egger's liner regression test were conducted in this meta-analysis. The P values of Egger's test were, respectively, .789, .262, and .267 for AFP, AFP-L3%, and their combination, indicating no evidence of a significant publication bias in this meta-analysis ( Figure 5).

| D ISCUSS I ON
The recommended noninvasive methods of HCC include radiographic techniques and the serum biomarkers in current. 24,25 Although AFP is one of the most widely applied tumor markers for  Item  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16 17  AFP and AFP-L3% for HCC diagnosis based an original study and literature review.
As most of markers present defective sensitivity, numerous studies indicate that it may be advisable to apply several biomarkers in subjects with HCC. 7,8 In the present meta-analysis, the overall sensitivity, specificity, and AUC of AFP were 0.59, 0.83, and 0.7322; those of AFP- The results of our meta-analysis are similar to these results in that the sensitivity of AFP + AFP-L3% was superior than that of AFP or AFP-L3% alone but the AUC value of AFP + AFP-L3% was inferior to their alone in HCC diagnosing. The present study has the following advantages over previous studies: First, more number of recent articles were included in our meta-analysis; second, Spearman analysis and meta-regression, involving three factors (test methodology, country of the first author, and sample size) were used for exploring the heterogeneity; third, all HCC patients in studies incorporated in our meta-analysis were diagnosed using the gold standard to prove the reliability of the primordial literature; fourth, we conducted an original study including 196 individuals to assess the diagnostic efficiency of the candidate markers.
All the results in our meta-analysis showed significant heterogeneity (all of I 2 > 50%). Although we made an effort to explore the heterogeneity using threshold effect analysis and meta-regression, none of the factors we analyzed was found to contribute to the high heterogeneity of the study. Thus, we conclude that certain other factors were responsible for the heterogeneity.
In the original study, the sensitivity and specificity were 0.592 and 0.918 for AFP, 0.367 and 1.000 for AFP-L3%, and 0.592 and 0.918 for AFP + AFP-L3%, respectively. The results above show that the ability of AFP-L3% is not beneficial to AFP, which is in agreement with the results of previous studies 28,29 ; moreover, the use of AFP in combination with AFP-L3% did not enhance the accuracy of distinguishing between subjects with and without HCC. Several factors may contribute to these results. First, our original study was a single-center, retrospective study. Second, the pathogenesis of included HCC patients are various (HBV, HCV, parasitization, et al.).
Third, the sample size is comparatively small.
In conclusion, our meta-analysis showed the diagnostic importance of AFP + AFP-L3% in terms of significantly higher sensitivity compared to that of either AFP or AFP-L3% alone in HCC diagnosis.
Moreover, the performance of AFP-L3% was greater to AFP for HCC F I G U R E 5 Begg's funnel plot. A, AFP, B, AFP-L3%, C, AFP + AFP-L3% diagnosis. Nevertheless, the results of our original study cannot validate absolutely that in our meta-analysis. Not only that, owing to the heterogeneity and various study limitations, further comprehensive research studies on a larger sample size are warranted to verify these findings.

AUTHOR S' CONTRIBUTIONS
ZW and JL conceived and designed the experiments; SC and XT performed the experiments; QX, YM, and HQ analyzed the data; LZ and LM contributed reagents/materials/analysis tools; and SC contributed to the writing of the manuscript.

DATA AVA I L A B I L I T Y S TAT E M E N T
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.