Diagnostic value of alpha-fetoprotein and prothrombin induced by vitamin K absence-II in serum, bile, and feces in hepatocellular carcinoma

Abstract

BACKGROUND

Hepatocellular carcinoma (HCC) is the most common pathological type of liver cancer and was the third leading cause of cancer-related deaths worldwide in 2020.

AIM

To evaluate the diagnostic potential of key tumor markers in serum, bile, and fecal samples for detecting HCC.

METHODS

Blood, bile, and fecal samples were collected from patients (n = 265) with HCC and cholecystitis from Guangxi Medical University’s First Affiliated Hospital. Immunohistochemistry was performed on 69 HCC samples, and 16S ribosomal RNA sequencing was conducted on 166 fecal samples. Tumor marker cut-off values in bile and feces were determined using the Youden index, while serum biomarkers followed hospital standards. Diagnostic performance was evaluated using receiver operating characteristic analysis.

RESULTS

The areas under the curve (AUCs) for distinguishing HCC were 0.898, 0.904, and 0.859 for serum alpha-fetoprotein (AFP), prothrombin induced by vitamin K absence-II (PIVKA-II), and bile AFP, respectively. Serum AFP had the highest diagnostic value (80%) for early-stage HCC. Combination analysis found that bile AFP and serum PIVKA-II achieved the highest AUC of 0.965 (P < 0.001), suggesting that bile AFP may serve as a valuable complementary biomarker, particularly in cases where serum AFP is not significantly elevated. Additionally, bile AFP was positively correlated with Actinomyces, which plays a significant role in promoting tumorigenesis; and was negatively correlated with Faecalibacterium, which was associated with robust anticancer immune responses (P < 0.05). These findings suggest the potential role of gut microbiota in modulating AFP levels and HCC progression.

CONCLUSION

Bile AFP improved the sensitivity of HCC detection, with the combination of bile AFP and PIVKA-II demonstrating the highest AUC for HCC diagnosis. AFP is associated with poorer clinical outcomes.

Key Words: Hepatocellular carcinoma; Alpha-fetoprotein; Prothrombin induced by vitamin K absence-II; Bile biomarkers; Serum biomarkers; Diagnosis; Receiver operating characteristic analysis; Gut-liver axis

Core Tip: This study evaluated the diagnostic potential of various biomarkers for hepatocellular carcinoma (HCC), with serum prothrombin induced by vitamin K absence-II (PIVKA-II) showing the highest diagnostic accuracy. However, serum alpha-fetoprotein (AFP) was more accurate for early HCC diagnosis. Bile AFP improved diagnostic performance and its combination with PIVKA-II achieved the highest area under the curve for HCC detection. Additionally, bile AFP was associated with poorer HCC clinical outcomes.

INTRODUCTION

Hepatocellular carcinoma (HCC) is the most common pathological type of liver cancer (LC)[1]. In 2020, HCC became the third leading cause of cancer-related deaths worldwide, a sharp rise from its rank as the ninth in 2000[2]. Consequently, most patients are diagnosed at an advanced stage with a poor prognosis, highlighting the urgent need for improved screening and early detection strategies[3,4].

Evidence suggests that tumor markers can increase 0.5-1.5 years before HCC becomes detectable on imaging[5]. Typically, biomarkers including alpha-fetoprotein (AFP), prothrombin induced by vitamin K absence-II (PIVKA-II)[6], and AFP-L3 are used for HCC detection[7]. Additionally, the gut microbiome may influence the progression of HCC through the gut-liver axis, potentially altering the composition of bile and fecal samples[8]. Other tumor markers such as carcinoembryonic antigen (CEA), cancer antigen 125 (CA125), CA153, and CA199 have also been associated with various gastrointestinal cancers. Unlike previous studies that primarily focused on serum biomarkers, the present study comprehensively evaluated multiple biomarkers in serum, bile, and fecal samples. Furthermore, since AFP expression promotes HCC development, microbiome analysis was combined with biomarker profiling to explore how the gut microbiota (GM) influences tumor markers and HCC progression. It is anticipated that the proposed innovative and comprehensive strategy for HCC detection will address the critical limitations of current diagnostic methods.

MATERIALS AND METHODS

Patient enrollment

From August 2018 to August 2020, a total of 265 eligible patients were admitted to the Department of Hepatobiliary Surgery at the First Affiliated Hospital of Guangxi Medical University (Nanning, China). Among them, 225 were diagnosed with HCC and 40 had chronic cholecystitis. This study was ethically approved by the hospital’s ethics committee (Approval No. 2020 [No. KY-E-118]). All patients voluntarily participated and provided written informed consent.

Inclusion criteria for the HCC group were as follows: (1) Patients between 18 and 80 years of age; (2) Patients with a diagnosis of HCC confirmed by postoperative pathology; and (3) Patients who have not undergone any previous treatments for HCC. Inclusion criteria for the control group were as follows: (1) Patients between 18 and 80 years old; and (2) Patients with a diagnosis of chronic cholecystitis confirmed by postoperative pathology. Patients with cholecystitis who had intrahepatic or extrahepatic bile duct stones, hepatic hemangioma, or liver function grade B-C.

HCC staging was determined using both the Barcelona clinic LC (BCLC) staging system and the China LC (CNLC) staging system[9,10].

Samples collection

Serum samples were obtained from all 265 patients before surgery, bile samples were collected from 118 patients during surgery, and fecal samples were collected from 166 patients before surgery. Fecal supernatant collection was performed with strict precision. Patients were instructed to collect midsection feces in a sterile 50 mL centrifuge tube, which was then stored at -80 °C until analysis. To prepare the fecal supernatant, a 5 g fecal sample was mixed with high-pressure distilled phosphate-buffered saline at a 1 g: 1 mL ratio. The mixture was centrifuged at 12000 × g for 5 minutes, and 5 mL of the supernatant was carefully extracted and filtered through a 40 μm filter. The filtrate was then transferred to a sterile blood collection tube and promptly submitted for analysis. All steps were meticulously performed to ensure the quality and integrity of the samples.

Laboratory tests

CEA, CA125, CA153, and CA199 are common tumor markers associated with various gastrointestinal cancers, while AFP and PIVKA-II are widely used for the diagnosis of HCC. Therefore, all samples were analyzed for AFP, PIVKA-II, CEA, CA125, CA153, and CA199 using a chemiluminescence microparticle immunoassay.

Immunohistochemical analysis of AFP in HCC

The main reagents included the EP209 alpha-1-fetoprotein monoclonal antibody (ZSGB-BIO, Beijing, China), cytokeratin 7 monoclonal antibody, clone UMAB161 (ZSGB-BIO), and polymerized horseradish peroxidase-anti MS/Rb IgG SD3003 (Celnovte Biotechnology, Rockville, MD, United States).

The experimental procedures are detailed in the Supplementary Methods section, available online. Immunohistochemical analyses were performed by expert pathologists in our department, who were blinded to the patients’ clinical and pathological information. A marker was considered positive if more than 10% of tumor cells exhibited moderate or intense AFP expression. By contrast, it was considered negative if no protein markers were present or if AFP expression was observed in less than 10% of tumor cells.

Detection of GM

DNA was extracted from the fecal samples of all patients. The extracted DNA was then amplified, amplicon libraries were constructed, and sequencing was performed using the Illumina MiSeq Platform (AIage Life Science Co. Ltd., Guangxi, China). Raw sequencing data underwent quality control and taxonomic annotation. The experimental methods followed those described by Pang et al[11], who explored the correlation between gut microbiome composition and the longevity of centenarians. Detailed procedures are provided in the online supplementary methods.

Statistical analysis

Statistical analysis was performed using SPSS version 22.0. Quantitative variables are expressed as the mean ± SD, while categorical variables are presented as proportions. For normally distributed data, comparisons were made using the Student’s t-test, whereas non-normally distributed data were analyzed using the Wilcoxon rank-sum test. Correlations between different biomarkers were assessed using Spearman’s rank correlation test.

Binary logistic regression analysis was performed to determine the coefficient (β) of each marker. The combined predictor of two markers was calculated using the following formula[12,13], where β1 and β2 are the regression coefficients for marker 1 and marker 2, respectively. This combined predictor enhances diagnostic accuracy by integrating both markers.

The optimal cut-off value for each biomarker was determined using Youden’s index[14], a metric that reflects both sensitivity and specificity. For bile and fecal biomarkers, the optimal cut-off values were selected based on the maximum Youden’s index, whereas hospital-standard cut-off values were used for serum biomarkers. Receiver operating characteristic (ROC) curves were plotted to evaluate the diagnostic performance of AFP and PIVKA-II, with area under the curve (AUC) values calculated[15,16]. AUC measures the discriminative ability of a biomarker, with higher values indicating better diagnostic performance. Sensitivity and specificity for all six markers in HCC diagnosis were calculated. The Kappa consistency test and paired χ² test (McNemar test) were used to compare sensitivity and specificity between different groups[17,18]. Furthermore, Kaplan-Meier analysis with the log-rank tests was conducted to compare recurrence-free survival (RFS) between patients with HCC with positive and negative AFP immunohistochemistry results. Statistical significance was set at P < 0.05 for all analyses.

RESULTS

Demographic characteristics

The study included patients aged 18 to 80 years. There was no statistically significant difference in age between the HCC and cholecystitis groups (P > 0.05). The incidence of HCC was higher in males, whereas cholecystitis was more prevalent in females, with a statistically significant difference between groups (P < 0.001). The infection rate of hepatitis B virus (HBV) was significantly higher in the HCC group compared to the cholecystitis group (P < 0.001). No significant difference in body mass index was observed between the groups (P > 0.05). All patients in the HCC group were newly diagnosed and had not received prior treatment. The cholecystitis group included patients with gallstone- or polyp-induced cholecystitis but no liver diseases such as intrahepatic or extrahepatic bile duct stones or hepatic hemangiomas. Neither group had major comorbidities.

Among the 225 patients with HCC, 156 (69.3%) were classified as BCLC stage A. Besides, according to the CNLC staging system, 97 (43.1%) were classified as stage Ia and 80 (35.6%) as stage Ib, indicating that the study primarily focuses on early-stage HCC.

Analysis of bile samples revealed that among 103 specimens from patients with HCC, 5 (4.9%) were negative for serum AFP but positive for bile AFP, and 1 (1.0%) was negative for serum PIVKA-II but positive for bile PIVKA-II. These findings suggest that bile biomarker analysis may enhance the detection of early-stage HCC.

Analysis of the difference of AFP and PIVKA-II levels between HCC and control groups

Figure 1 illustrates significant differences in serum AFP, PIVKA-II levels, and bile AFP levels between the HCC and control groups (P < 0.001). These findings suggest that serum AFP, PIVKA-II, and bile AFP hold strong potential as diagnostic biomarkers for HCC.

Figure 1 Differences in tumor marker levels among sample types. A-F: Differences in tumor markers in blood, bile, and feces of patients with hepatocellular carcinoma (HCC) and non-HCC; G and H: Comparative analysis of alpha-fetoprotein (AFP) and prothrombin induced by vitamin K absence-II (PIVKA-II) levels in blood and bile among patients with HCC.

Analysis of clinicopathological data of patients with HCC

Among the 225 patients included in this study, 214 (95.1%) did not exhibit portal vein tumor thrombus, and 195 (86.7%) had only a single tumor. The maximum tumor diameter (d) was ≤ 5 cm in 145 (64.5%) cases. Additionally, distant metastasis was observed in only 4 (1.8%) patients. A total of 198 (88%) patients had moderate differentiation, and microvascular invasion (MVI) was absent (MVI = 0) in 169 (75.1%) cases. These pathological features suggest that the study predominantly focuses on early-stage patients with HCC.

Correlation between different biomarkers and different pathological characteristics

We conducted a correlation analysis using the Spearman correlation coefficient. The results revealed a significant positive correlation between serum AFP and bile AFP, with a correlation coefficient of 0.9 (P < 0.001). In addition, serum AFP, serum PIVKA-II, and bile AFP also showed significant positive correlations with tumor BCLC staging and MVI grade (P < 0.05), as shown in Supplementary Table 1.

ROC curve analysis of six markers in serum, bile, and fecal samples for the detection of HCC

CEA, CA125, CA153, and CA199 were not particularly effective at distinguishing between patients with HCC and non-HCC (P > 0.001). Similarly, all of the fecal supernatants were not particularly effective at distinguishing between patients with HCC and non-HCC counterparts (P > 0.001). The maximum AUCs for distinguishing HCC from chronic cholecystitis in serum were 0.898 for AFP and 0.904 for PIVKA-II. The maximum AUCs for distinguishing HCC from cholecystitis in bile were 0.8559 for AFP and 0.526 for PIVKA-II. Additionally, serum PIVKA-II had the highest AUC for distinguishing HCC from cholecystitis in all samples (P < 0.001) (Table 1 and Figure 2A-C).

Figure 2 Diagnostic efficacy, correlation coefficients of tumor markers, immunohistochemical analysis, and recurrence-free survival analysis. A: Receiver operating characteristic (ROC) curve for hepatocellular carcinoma (HCC) diagnosis using alpha-fetoprotein (AFP) or prothrombin induced by vitamin K absence-II (PIVKA-II) in serum, bile, and feces; B: ROC curve for HCC diagnosis using the combination of AFP and PIVKA-II in serum and bile; C: ROC curve for the HCC diagnosis using carcinoembryonic antigen (CEA), carbohydrate antigen 125 (CA125), CA153, and CA19-9 in serum; D: Correlation coefficient between serum AFP, PIVKA-II, and bile AFP; E: Correlation analysis of AFP expression in tumor tissues and serum AFP levels; F: Correlation analysis between AFP expression in tumor tissues and bile AFP levels; G: Correlation analysis between AFP expression in paracancerous tissue and serum AFP levels; H: Correlation analysis between AFP expression in paracancerous tissue and bile AFP levels; I: Recurrence-free survival analysis using AFP immunohistochemistry in tumor tissue for serum in patients with AFP-positive HCC; J: Recurrence-free survival analysis using AFP immunohistochemistry in paracancerous tissue for serum in patients with AFP-positive HCC. AUC: Area under the curve; CI: Confidence interval.

Table 1 Analysis of the diagnostic value of alpha-fetoprotein and prothrombin induced by vitamin K absence-II in serum, bile, and feces in hepatocellular carcinoma.

Samples	Indicators	Cut-off value	Sensitivity (%)	Specificity (%)	AUC (95%CI)	P value
Serum	AFP, ng/mL	8.78	65.8	100.0	0.898 (0.859-0.938)	< 0.001
	PIVKA-II, mAU/mL	40.00	78.2	90.9	0.904 (0.866-0.942)	< 0.001
	CEA, ng/mL	5.00	10.5	100.0	0.621 (0.459-0.782)	0.159
	CA125, U/mL	35.00	7.6	66.7	0.506 (0.298-0.715)	0.940
	CA153, U/mL	31.3	1.9	100.0	0.628 (0.464-0.791)	0.137
	CA199, U/mL	37.00	11.4	100.0	0.604 (0.433-0.775)	0.225
Bile	AFP, ng/mL	1.32	61.2	100.0	0.859 (0.779-0.938)	< 0.001
	PIVKA-II, mAU/mL	59.50	35.9	80.0	0.526 (0.384-0.668)	0.725
	CEA, ng/mL	84.26	9.4	100.0	0.371 (0.179-0.563)	0.219
	CA125, U/mL	123.75	3.8	100.0	0.211 (0.048-0.374)	0.006
	CA153, U/mL	1.35	56.6	66.7	0.596 (0.407-0.786)	0.358
	CA199, U/mL	4.31	92.3	44.4	0.622 (0.394-0.850)	0.246
Feces	Fecal AFP, ng/mL	0.04	16.4	96.2	0.568 (0.452-0.684)	0.272
	Fecal PIVKA-II, mAU/mL	1469.20	5.7	96.2	0.402 (0.285-0.519)	0.115
	CEA, ng/mL	15001.00	0.0	100.0	0.431 (0.156-0.706)	0.659
	CA125, U/mL	23.25	6.9	100.0	0.297 (0.000-0.611)	0.195
	CA153, U/mL	1.85	17.2	100.0	0.522 (0.229-0.814)	0.890
	CA199, U/mL	8.64	62.1	75.0	0.575 (0.462-0.689)	0.659
Combination	Fecal AFP + fecal PIVKA-II	-0.0016	56.4	77.0	0.631 (0.514-0.748)	0.034
	Serum AFP + bile AFP	9.34	70.0	100.0	0.903 (0.843-0.964)	< 0.001
	Serum AFP + serum PIVKA-II	10.84	90.7	97.0	0.961 (0.940-0.983)	< 0.001
	Serum PIVKA-II + bile AFP	113.30	88.3	100.0	0.965 (0.933-0.996)	< 0.001
	Serum PIVKA-II + bile PIVKA-II	29.90	86.3	92.3	0.913 (0.860-0.967)	< 0.001
	Serum AFP + bile PIVKA-II	4.35	81.4	92.3	0.891 (0.827-0.956)	< 0.001
	Bile AFP+ bile PIVKA-II	1.09	60.2	100.0	0.871 (0.794-0.947)	< 0.001

AFP: Alpha-fetoprotein; AUC: Area under the curve; CA: Carbohydrate antigen; CEA: Carcinoembryonic antigen; CI: Confidence interval; PIVKA-II: Prothrombin induced by vitamin K absence-II.

Regarding combinations of biomarkers, the combination of “bile AFP and serum PIVKA-II” showed the highest AUC of 0.965 (P < 0.001) compared to the other combinations (Table 1 and Figure 2B). Additionally, serum AFP and bile AFP exhibited a strong correlation (r = 0.9, P < 0.001; Figure 2D). These findings collectively suggest that bile AFP may be useful for differentiating between patients with HCC and non-HCC.

Sensitivity and specificity of six markers for the diagnosis of HCC

The sensitivities, specificities, and cut-off values of the six independent markers for all patients are shown in Table 1. In bile samples, AFP had a sensitivity of 61.2% with a cut-off of 1.32 ng/mL, whereas in serum samples, it had a sensitivity of 65.8% at a cut-off value of 8.78 ng/mL. PIVKA-II, with a cut-off value of 40 mAU/mL, demonstrated an increased sensitivity of 78.2%. Table 2 shows that bile AFP had a diagnostic efficacy of 4.9% (5 of 103 cases) in the serum AFP-negative but PIVKA-II-positive group and 10.7% (11 of 103 cases) in the serum PIVKA-II-negative but AFP-positive group.

Table 2 Analysis of the diagnostic efficacy of alpha-fetoprotein and prothrombin induced by vitamin K absence-II in serum alpha-fetoprotein or prothrombin induced by vitamin K absence-II negative group.

Group	Indicators	Sensitivity (%)	AUC (95%CI)	P value
Serum AFP negative group	Serum PIVKA-II	75.3	0.887 (0.827-0.947)	< 0.001
	Bile AFP	13.9	0.657 (0.487-0.828)	0.079
	Bile PIVKA-II	38.9	0.537 (0.369-0.705)	0.679
	Fecal AFP	0.0	0.512 (0.347-0.649)	0.868
	Fecal PIVKA-II	4.3	0.386 (0.250-0.521)	0.105
Serum PIVKA-II negative group	Serum AFP	61.2	0.848 (0.765-0.931)	< 0.001
	Bile AFP	55.0	0.842 (0.714-0.969)	0.001
	Bile PIVKA-II	5.0	0.270 (0.088-0.452)	0.021
	Fecal AFP	6.0	0.561 (0.415-0.708)	0.418
	Fecal PIVKA-II	6.0	0.352 (0.214-0.491)	0.051
Both serum AFP and serum PIVKA-II negative group	Bile AFP	0.0	0.648 (0.426-0.871)	0.233
	Bile PIVKA-II	11.1	0.230 (0.018-0.441)	0.030
	Fecal AFP	18.2	0.483 (0.274-0.692)	0.872
	Fecal PIVKA-II	9.0	0.414 (0.209-0.619)	0.412

AFP: Alpha-fetoprotein; AUC: Area under the curve; CI: Confidence interval; PIVKA-II: Prothrombin induced by vitamin K absence-II.

In addition, to enhance the diagnostic value for HCC detection, we combined AFP and PIVKA-II and assessed the sensitivities and specificities of different combinations (Table 1). The combination of AFP and PIVKA-II in serum yielded the highest sensitivity of 90.7% and a specificity of 97%. The combination of bile AFP and serum PIVKA-II achieved a sensitivity of 88.3% and the highest specificity of 100%.

We used the Kappa consistency test and paired χ² test to compare the sensitivity and specificity of the combined markers. The combination of “AFP in bile and PIVKA-II in serum” demonstrated significantly higher sensitivity compared to serum PIVKA-II (P = 0.035) or bile AFP alone (P < 0.001) for HCC detection.

Analysis of the diagnostic value of AFP and PIVKA-II in different subgroups of serum, bile, and feces to HCC

As shown in Table 3, the HCC population was divided into subgroups based on various staging systems. According to the BCLC staging system, the HCC population was categorized into four subgroups: BCLC 0, BCLC A, BCLC B, and BCLC C + D. The CNLC staging system further divided the HCC population into four subgroups: CNLC Ia, CNLC Ib, CNLC II, and CNLC III + IV. Based on tumor diameter, the HCC population was categorized into four subgroups: “Tumor diameter ≤ 2 cm”, “2 cm < tumor diameter ≤ 5 cm”, “5 cm < tumor diameter ≤ 10 cm”, and “tumor diameter > 10 cm”. According to the number of tumors, the HCC population was divided into two subgroups: “single tumor” and “multiple tumors”.

Table 3 Analysis of the diagnostic efficacy of alpha-fetoprotein and prothrombin induced by vitamin K absence-II for hepatocellular carcinoma in different tumor stages, size and number subgroups.

Characteristic		Serum			Bile			Fecal supernatant
		Sensitivity (%)	AUC (95%CI)	P value	Sensitivity (%)	AUC (95%CI)	P value	Sensitivity (%)	AUC (95%CI)	P value
Tumor staging (CNLC)
Ia	AFP, ng/mL	65.1	0.900 (0.848-0.952)	< 0.001	46.5	0.805 (0.688-0.923)	< 0.001	11.1	0.555 (0.424-0.686)	0.419
	PIVKA-II, mAU/mL	63.5	0.835 (0.768-0.903)	< 0.001	18.6	0.383 (0.209-0.557)	0.180	12.7	0.466 (0.337-0.596)	0.620
Ib	AFP, ng/mL	63.8	0.898 (0.842-0.954)	< 0.001	67.5	0.882 (0.797-0.968)	< 0.001	18.8	0.571 (0.436-0.706)	0.316
	PIVKA-II, mAU/mL	91.3	0.960 (0.925-0.994)	< 0.001	41.9	0.609 (0.458-0.761)	0.211	0.0	0.333 (0.202-0.463)	0.018
II	AFP, ng/mL	71.0	0.878 (0.777-0.978)	< 0.001	66.7	0.874 (0.728-1)	0.003	27.3	0.607 (0.443-0.770)	0.207
	PIVKA-II, mAU/mL	80.6	0.929 (0.858-0.999)	< 0.001	55.6	0.652 (0.418-0.886)	0.222	0.0	0.392 (0.231-0.552)	0.200
III + IV	AFP, ng/mL	88.2	0.929 (0.818-1)	< 0.001	100.0	1.00 (1-1)	< 0.001	14.3	0.544 (0.296-0.792)	0.725
	PIVKA-II, mAU/mL	94.1	0.989 (0.966-1)	< 0.001	75.0	0.708 (0.453-0.963)	0.107	0.0	0.342 (0.095-0.587)	0.202
Tumor staging (BCLC)
0	AFP, ng/mL	80.0	0.924 (0.817-1)	< 0.001	50.0	0.850 (0.649-1)	0.036	16.7	0.644 (0.401-0.887)	0.277
	PIVKA-II, mAU/mL	40.0	0.679 (0.448-0.910)	0.09	0.0	0.267 (0.047-0.486)	0.162	50.0	0.622 (0.3-0.944)	0.359
A	AFP, ng/mL	59.0	0.884 (0.836-0.932)	< 0.001	57.3	0.837 (0.747-0.927)	< 0.001	15.5	0.560 (0.439-0.680)	0.351
	PIVKA-II, mAU/mL	77.6	0.900 (0.856-0.943)	< 0.001	33.3	0.505 (0.358-0.652)	0.953	4.1	0.398 (0.275-0.521)	0.111
B	AFP, ng/mL	77.1	0.926 (0.852-1)	< 0.001	64.3	0.900 (0.792-1)	< 0.001	21.7	0.614 (0.454-0.773)	0.173
	PIVKA-II, mAU/mL	82.9	0.932 (0.868-0.997)	< 0.001	28.6	0.557 (0.344-0.770)	0.600	4.3	0.390 (0.23-0.549)	0.186
C + D	AFP, ng/mL	87.5	0.941 (0.859-1)	< 0.001	90.0	0.967 (0.897-1)	< 0.001	0.0	0.516 (0.324-0.709)	0.865
	PIVKA-II, mAU/mL	91.7	0.985 (0.961-1)	< 0.001	80.0	0.747 (0.524-0.970)	0.040	0.0	0.360 (0.181-0.538)	0.148
Tumor diameter (cm)
d ≤ 2	AFP, ng/mL	73.8	0.898 (0.822-0.937)	< 0.001	62.5	0.817 (0.663-0.970)	0.003	16.7	0.547 (0.395-0.699)	0.549
	PIVKA-II, mAU/mL	52.4	0.792 (0.691-0.893)	< 0.001	43.8	0.575 (0.369-0.781)	0.477	6.7	0.347 (0.201-0.494)	0.051
2< d ≤ 5	AFP, ng/mL	60.2	0.912 (0.863-0.961)	< 0.001	57.8	0.841 (0.740-0.941)	< 0.001	12.3	0.581 (0.453-0.709)	0.223
	PIVKA-II, mAU/mL	75.7	0.887 (0.834-0.941)	< 0.001	33.3	0.542 (0.382-0.702)	0.627	6.2	0.435 (0.303-0.567)	0.330
5 < d ≤ 10	AFP, ng/mL	67.3	0.884 (0.814-0.955)	< 0.001	58.8	0.88 (0.785-0.976)	< 0.001	0.0	0.502 (0.346-0.658)	0.979
	PIVKA-II, mAU/mL	94.5	0.980 (0.952-1)	< 0.001	35.3	0.486 (0.311-0.661)	0.879	7.1	0.371 (0.222-0.520)	0.104
d > 10	AFP, ng/mL	72.0	0.874 (0.760-0.988)	< 0.001	87.5	0.950 (0.849-1)	< 0.001	35.3	0.689 (0.52-0.858)	0.038
	PIVKA-II, mAU/mL	96.0	0.993 (0.977-1)	< 0.001	37.5	0.508 (0.224-0.793)	0.949	0.0	0.362 (0.188-0.536)	0.130
Number of tumors
Single	AFP, ng/mL	65.6	0.901 (0.86-0.9441)	< 0.001	58.9	0.849 (0.765-0.934)	< 0.001	14.0	0.555 (0.435-0.674)	0.381
	PIVKA-II, mAU/mL	77.4	0.896 (0.856-0.937)	< 0.001	36.7	0.539 (0.394-0.683)	0.634	6.6	0.409 (0.289-0.529)	0.148
Multiple	AFP, ng/mL	66.7	0.883 (0.785-0.980)	< 0.001	76.9	0.923 (0.825-1)	< 0.001	31.6	0.651 (0.483-0.818)	0.087
	PIVKA-II, mAU/mL	83.3	0.954 (0.910-0.997)	< 0.001	30.8	0.441 (0.215-0.667)	0.596	0.0	0.358 (0.191-0.526)	0.108

AFP: Alpha-fetoprotein; AUC: Area under the curve; BCLC: Barcelona Clinic Liver Cancer staging system; CI: Confidence interval; CNLC: China liver cancer staging system; PIVKA-II: Prothrombin induced by vitamin K absence-II.

In BCLC stage 0, CNLC stage Ia, and HCC with a tumor diameter ≤ 2 cm, the diagnostic sensitivity of serum AFP (80%, 73.8%) was higher than that of serum PIVKA-II (40%, 52.4%). Bile AFP was significantly more effective than serum AFP in diagnosing multiple HCCs (sensitivity 87.5%, AUC = 0.950) or HCCs with a diameter > 10 cm (sensitivity 76.9%, AUC = 0.923) (P < 0.001).

Expression of AFP in LC tissues and paracancerous tissues

Of the 72 LC tissue samples, 17 tested positive for AFP, achieving a positivity rate of 23.6%. By contrast, only 6 of 72 paraneoplastic tissue samples were positive for AFP. The positivity rate of AFP was significantly higher in LC tissues compared to paracancerous tissues, and the staining was primarily localized in the cytoplasm of the cells, as shown in Figure 3A and B. By contrast, a negative AFP immunohistochemistry result is displayed in Figure 3C. Elevated levels of serum or bile AFP (including the serum AFP-negative but bile AFP-positive group) were observed in AFP immunohistochemistry-positive patients. Further analysis revealed that the rates of positive AFP immunohistochemistry in tumor tissues were significantly correlated with serum (r = 0.686, P < 0.001) and bile (r = 0.672, P < 0.001) AFP levels, in contrast, this correlation was markedly weaker in paracancerous tissues, as depicted in Figure 2E-H.

Figure 3 Immunohistochemical analysis of alpha-fetoprotein in hepatocellular carcinoma and the structural relationship between alpha-fetoprotein and the spatial distribution of bile ducts. A: Alpha-fetoprotein (AFP) immunohistochemistry in tumor tissue; B: AFP immunohistochemistry in paraneoplastic tissue; C: Negative AFP immunohistochemistry result in tumor tissue, with scattered enhancement in some areas; D: AFP immunohistochemistry results; E: Cytokeratin 7 immunohistochemistry results. A-C from left to right: 100 ×, 200 ×, 400 × magnifications; D and E from left to right: 100 ×, 100 ×, 200 × magnifications.

To explore the presence of bile AFP in patients with serum AFP-negative HCC and to investigate the spatial distribution of AFP in tumor tissue relative to bile ducts, we used tissue sections adjacent to the tumor. Bile duct epithelial cells were labeled using cytokeratin 7 monoclonal antibody and compared with adjacent tumor tissues previously stained for AFP (Figure 3D-E).

Further analysis revealed that there was no significant difference in RFS between AFP immunohistochemistry-negative and AFP immunohistochemistry-positive patients with HCC in tumor tissues (P = 0.869, hazard ratio [HR] = 1.085, 95% confidence interval [CI]: 0.4069-2.893) and adjacent tissues (P = 0.102, HR = 4.584, 95%CI: 1.441-14.58), as demonstrated in Figure 2I and J.

Correlation analysis of GM and tumor markers

To investigate the relationship between the gut microbiome and tumor markers in patients with HCC, we identified the top 25 bacterial genera that showed a significant correlation with the tumor markers, as represented in Figure 4.

Figure 4 Correlation analysis of gut microbiota and tumor markers. ^aP < 0.05. ^bP < 0.01. ^cP < 0.001. AFP: Alpha-fetoprotein; PIVKA-II: Prothrombin induced by vitamin K absence-II.

Serum AFP was positively correlated with Coprococcus 1, Ruminococcus gauvreauii group, Eubacterium hallii group, Lachnospiraceae ND3007 group, Faecalibacterium, and Actinomyces (P < 0.05). Serum PIVKA-II was negatively correlated with Tyzzerella 3, Moryella, and Lachnoclostridium (P < 0.05). Bile AFP was positively correlated with Actinomyces and negatively correlated with Faecalibacterium (P < 0.05). Bile PIVKA-II was positively correlated with the Eubacterium eligens group (P < 0.05), as demonstrated in Figure 4.

DISCUSSION

This study evaluated the diagnostic potential of various biomarkers for HCC, including CEA, CA125, CA153, CA199, AFP, and PIVKA-II in serum, bile, and fecal samples. CEA is widely used in the detection of colorectal cancer but its utility for early-stage detection is limited. Its primary advantage lies in its strong association with poor prognosis and its high sensitivity in detecting recurrent colorectal cancer, providing an approximate lead time of 5 months[19]. As a marker for the early diagnosis of ovarian cancer in high-risk populations, CA125 is used for diagnosing ascites in patients with liver cirrhosis[20]. CA153 is a specific breast cancer marker, playing a significant role in its diagnosis and postoperative monitoring[21], although elevated levels can also be observed in other tumors[22]. CA199 is a sensitive marker for pancreatic cancer and has moderate sensitivity in gastrointestinal cancers including ovarian, colorectal, and gastric cancers[23-25]. Since fecal samples are commonly used to study the impact of microbiota on diseases, particularly colorectal cancer, fecal supernatants were innovatively used to detect tumor marker levels. However, CEA, CA125, CA153, and CA199 levels in the blood, bile, and feces were not effective in distinguishing patients with and without HCC, consistent with previous studies indicating that CA125 does not significantly distinguish HCC[26].

Compared to other tumor markers, CA125 is more closely associated with tumors with a diameter greater than 5 cm and a poorer Child-Pugh classification, especially in those patients with ascites. However, its diagnostic value for early-stage HCC remains limited[27]. Yang et al’s study suggests that CA153 might have some diagnostic value for LC but it was significantly elevated in the advanced disease stages[26] and is primarily associated with liver metastasis of breast cancer and ocular metastasis of LC[28,29]. In addition, it has been reported that CA199 has a diagnostic sensitivity of 64.28%, while CEA has a sensitivity of 83.67% in HCC[30]. However, since most of our participants were in the early HCC stage, these markers had limited utility in distinguishing HCC.

As one of the earliest tumor markers identified, AFP was initially recognized as a serum marker for HCC and later found to be associated with germ cell tumors[31]. Our study showed that serum AFP had a sensitivity of 65.8% and a specificity of 100% for diagnosing HCC, aligning with the findings of Wong et al[32] who reported a sensitivity range of 39% to 65% and a specificity range of 76% to 94% for serum AFP. Moreover, previous studies have shown that AFP levels do not significantly increase in 80% of small LCs[33,34]. However, in contrast to previous studies, the sensitivity of serum AFP was higher, reaching 73.8% for HCC tumors smaller than 2 cm. Since AFP can also be elevated in conditions such as active hepatitis, liver cirrhosis, and certain non-HCC tumors[35,36], it is crucial to incorporate additional tumor markers to enhance its diagnostic value.

PIVKA-II is a diagnostic marker for HCC[37], with a reported sensitivity of 62% and a specificity of 95% for diagnosing HCC using a cut-off value of 40 mAU/mL[32]. In the present study, serum PIVKA-II showed a sensitivity of 78.2% and specificity of 90.9% when using the same cut-off value, higher than those reported in previous studies. Furthermore, the sensitivity of PIVKA-II was superior to that of AFP, highlighting the improved diagnostic performance of PIVKA-II in this study. It has been reported that AFP has an advantage over PIVKA-II in diagnosing early-stage LC, particularly when tumors are smaller than 3 cm or 2 cm[36,38], but the overall diagnostic value of PIVKA-II was superior to that of AFP in the present study. However, AFP levels may begin to rise as early as 6 months before an HCC diagnosis[39], explaining why serum AFP was more effective than serum PIVKA-II for detecting early-stage HCC.

Emerging research has highlighted the release of extracellular vesicles (EVs) by malignant tumors into surrounding fluids, which has shown promise in cancer diagnosis. Indeed, bile EVs could be used to differentiate between malignant and non-malignant common bile duct strictures with 63.3% accuracy[24]. Also, CA125, either alone or combined with CEA, could effectively distinguish between benign and malignant causes of biliary obstruction[40]. In the present study, bile AFP could diagnose 61.2% of patients with HCC, with even higher sensitivity in cases of tumors smaller than 2 cm, indicating the diagnostic potential of bile for LC. AFP-producing human cholangiocarcinoma cells exhibited enhanced tumorigenic and malignant potential in vitro and in vivo[41]. Interestingly, although many AFP-enriched hepatocytes were identified, they were not located near the bile ducts, consistent with our focus on HCC rather than cholangiocarcinoma. Additionally, bile AFP levels were positively correlated with AFP expression in tumor tissues, leading us to hypothesize that AFP produced by hepatocytes is secreted into the bile via the canalicular system.

The combination of serum AFP with bile AFP increased sensitivity from 65.8% to 70%. Although serum AFP combined with serum PIVKA-II showed the highest sensitivity for diagnosing HCC (90.7%), serum PIVKA-II combined with bile AFP had the highest AUC of 0.965. This suggests that bile AFP, when combined with serum PIVKA-II, can significantly improve the diagnostic sensitivity for differentiating patients with and without HCC, particularly in cases where AFP levels alone are insufficient for diagnosis. The improved AUC value indicates a potential clinical advantage as it could contribute to earlier detection and more accurate HCC diagnosis, especially in high-risk populations. Given the critical importance of early detection in improving patient outcomes, these combined biomarkers present a promising diagnostic strategy that could complement current clinical methods, offering more reliable results in challenging diagnostic situations.

There was also a strong correlation between AFP levels and cancer tissues, particularly in cases with poor pathological types. It has been reported that AFP is more closely associated with poorly differentiated LC[41] and can enhance human LC cell proliferation[42]. When AFP expression was silenced in the HCC cell line Huh7, cell proliferation was inhibited by 46.15%[43]. In addition, the GM can influence the tumor immune microenvironment and contribute to the development of LC through the gut-liver axis[44,45]. AFP is associated with tumor malignancy and the GM composition is linked to tumor formation. In the present study, there was a strong negative correlation between bile AFP and Faecalibacterium as well as a strong positive correlation with Actinomyces. Previous studies have shown that Faecalibacterium is associated with robust anticancer immune responses in patients with cancer[46], while Actinomyces plays a significant role in promoting tumorigenesis[47]. These findings confirm the relationship among the GM, tumor markers, and tumor malignancy but further studies are required to determine the underlying mechanisms. Although AFP correlates with HCC malignancy, our immunohistochemical analysis did not reveal a significant correlation between AFP positivity in tumor tissues and patient RFS. However, we suggest that patients with HCC with positive serum and bile AFP levels should be closely monitored for potential tumor recurrence.

Our findings have significant implications for the early diagnosis and clinical management of HCC. First, since serum AFP demonstrated a sensitivity of 80% in early HCC diagnosis, outperforming PIVKA-II, it should be prioritized in HCC screening given its affordability and diagnostic efficiency. For high-risk populations, such as patients with chronic hepatitis B or liver cirrhosis, a combined test of AFP and PIVKA-II is recommended when feasible, as it enhances diagnostic accuracy and facilitates early detection. Second, the combination of serum PIVKA-II and bile AFP exceeded the diagnostic performance of AFP or PIVKA-II alone, suggesting that bile AFP testing could serve as a valuable supplementary tool, improving both sensitivity and specificity for patients with suspected HCC but without significantly elevated serum AFP. For patients undergoing biliary procedures such as endoscopic retrograde cholangiopancreatography or biliary drainage, bile AFP may provide additional diagnostic value, potentially increasing the early detection of HCC. Third, our study demonstrated a strong correlation between AFP levels and HCC histological subtypes, with notably high AFP expression in poorly differentiated HCC cases. Moreover, silencing AFP expression reduced HCC cell proliferation, supporting its role in HCC malignancy and progression. Thus, AFP not only serves as a diagnostic biomarker but also aids in assessing disease progression and recurrence risk and may serve as a potential therapeutic target. Finally, this study is the first to reveal a potential link between bile AFP levels and GM with bile AFP negatively correlated with Faecalibacterium and positively correlated with Actinomyces. Since Faecalibacterium is linked to anti-cancer immune responses, whereas Actinomyces contributes to tumorigenesis, the GM may influence the tumor microenvironment and regulate AFP expression and HCC progression but further studies are required to determine the underlying mechanisms.

The study had several limitations. First, most participants were infected with HBV, which may limit the generalizability of the findings to other etiological groups, such as those with cirrhosis related to hepatitis C, alcoholic or non-alcoholic steatohepatitis, or type 2 diabetes. Second, the control group consisted of patients with chronic cholecystitis rather than cirrhosis, mainly due to easier access to bile samples. Although this may introduce potential bias, it is unlikely to significantly affect the results. Third, since most HCC surgical patients were in the early stages of cancer, there might be potential for selection bias, meaning that the sampled population may not fully represent all patients with LC, which could impact the generalizability of our findings. Finally, our study population was primarily derived from the Guangxi region of China, and variations in ethnicity, geographic location, and clinical characteristics might lead to differences in outcomes. Therefore, the applicability of our findings to broader populations requires further validation with multicenter data. In the future, we plan to expand our clinical sample collection and collaborate in multicenter studies to enhance the reliability and generalizability of our research.

CONCLUSION

This study represents a significant advancement by applying classical tumor markers for HCC to unconventional specimens, such as bile and feces. Importantly, the combination of serum AFP, PIVKA-II, and bile AFP significantly enhanced HCC detection efficiency, with serum AFP proving to be a more effective diagnostic marker for early-stage HCC compared to serum PIVKA-II. Moreover, AFP was linked to the malignancy of HCC. Therefore, patients with dual positivity for serum and bile AFP, in particular, require enhanced monitoring to prevent tumor progression and recurrence.

ACKNOWLEDGEMENTS

We would like to express our gratitude to all those who contributed to the successful completion of this research. We extend our appreciation to the Key Laboratory of Early Prevention and Treatment for Regional High-Frequency Tumors (Guangxi Medical University)-Ministry of Education. Their cooperation was essential to the accomplishment of this work.