Abstract
The ASAP Model and GAAD Model integrate gender, age, AFP, and PIVKA-II to predict hepatocellular carcinoma (HCC) risk. This study compared their diagnostic performance in a high-risk cohort. A total of 352 subjects were enrolled, including 115 HCC patients, 137 with chronic liver disease (CLD), and 100 healthy controls (HCs). Serum AFP and PIVKA-II levels were measured using Abbott and Roche assays, respectively. ASAP and GAAD scores were calculated. Using manufacturer-recommended cutoffs and predefined thresholds (ASAP ≥ 33.4%, ASAP ≥ 66.7%, GAAD ≥ 2.57), we evaluated the diagnostic performance of each model. Area under the receiver operating characteristic curves (AUCs) were compared using the DeLong test. In the HCC group, both the median values and positive rates for ASAP and GAAD scores were significantly higher than those observed in the CLD group and HCs (P < 0.01). ASAP ≥ 33.4% showed the highest sensitivity (0.974), although its specificity was lower (0.658). ASAP ≥ 66.7% exhibited the highest specificity (0.894) and accuracy (0.901). However, neither threshold could definitively exclude HCC in low-risk ranges. For Test Cohort 1 (including the HCC group and other non-HCC groups), the AUCs for ASAP and GAAD scores were 0.955 and 0.958, respectively (P = 0.610). For Test Cohort 2 (including HCC and CLD groups), the AUCs for ASAP and GAAD scores were 0.926 and 0.928 (P = 0.810). For Test Cohort 3 (including the CLD and HCs groups), the AUCs for ASAP and GAAD scores were 0.833 and 0.832 (P = 0.944). Both ASAP and GAAD demonstrated excellent diagnostic performance for HCC, significantly outperforming single tumor markers. No significant differences were observed between the two models across all risk cohorts, suggesting their comparable clinical utility.
Subject terms: Biomarkers, Cancer, Diseases, Gastroenterology, Oncology, Risk factors
Introduction
According to data released by the National Cancer Center of China in 2022, the incidence of primary hepatocellular carcinoma (HCC) in China is 367,700, ranking 4th among all types of cancer in terms of new cases and 5th in incidence rate. In the same year, the number of deaths due to primary liver cancer reached 316,500, making it the second-leading cause of cancer deaths and second in mortality rate. Therefore, early detection of HCC in individuals at high risk is critical for diagnosis, treatment, and prognosis1.
High-risk groups for liver cancer in China primarily include individuals with hepatitis B virus (HBV) and/or hepatitis C virus (HCV) infections, excessive alcohol consumption, fatty liver degeneration or metabolic dysfunction-related liver diseases, dietary exposure to aflatoxin B1, cirrhosis from other causes, and those with a family history of liver cancer, particularly men over the age of 402.
Serum alpha-fetoprotein (AFP) is the most widely used biomarker for liver cancer screening, diagnosis, and monitoring. However, AFP levels may also be elevated in various chronic liver diseases (CLDs), pregnancy, gastrointestinal disorders, and other malignancies. Its sensitivity and specificity for diagnosing HCC are approximately 60% and 80%, respectively3,4. PIVKA-II, a protein induced by vitamin K deficiency or antagonist-II, has recently emerged as a promising novel marker for HCC detection, particularly in AFP-negative populations5. While both biomarkers have limitations in clinical applications, their combined use can enhance the detection of early-stage HCC5–7.
In 2019, Yang et al. introduced the ASAP score, which incorporates age, gender, AFP, and PIVKA-II, effectively improving diagnostic accuracy for primary liver cancer. This score can be applied to the screening and surveillance of high-risk groups8. The GAAD algorithm, proposed by Piratvisuth et al. in 2023, also integrates data on age, gender, AFP, and PIVKA-II. Both the ASAP and GAAD algorithms are simplified versions of the GALAD model, with AFP-L3 removed, and exhibit comparable diagnostic performance. These digitized algorithms provide valuable evidence for detecting early-stage HCC in patients with CLD under surveillance[9). While both algorithms serve as adjuncts in diagnosing primary liver cancer, a direct comparison of their characteristics, advantages, and specific performance has not been conducted in existing literature. Comparing the performance and utility of these algorithms in clinical and laboratory settings is a valuable direction for further research.
This study aims to evaluate the screening and diagnostic value of ARCHITECT AFP, ARCHITECT PIVKA-II, the ASAP score, Elecsys AFP, Elecsys PIVKA-II, and the GAAD model for HCC, using corresponding reagent kits and detection systems from two manufacturers. The application of these methods in screening and diagnosing various high-risk populations for HCC was compared, with the ultimate goal of identifying the optimal strategy for screening and diagnosis in high-risk groups.
Methods
Patient selection and study design
Patients and healthy controls (HCs) (aged ≥ 18 years) from the Sichuan Provncial People’s Hospital were recruited between October 2024 and January 2025 and categorized into three groups: (1) Primary HCC group: patients with a first-time diagnosis of HCC confirmed radiologically or by liver biopsy9,10; (2) CLD group, including liver cirrhosis, acute or chronic hepatitis, alcoholic liver disease, and non-alcoholic steatohepatitis, with positive AFP or PIVKA-II; (3) HCs group: individuals undergoing routine health checkups at the medical center, with no abdominal diseases identified by ultrasound.
Exclusion criteria were as follows: (1) patients with multiple organ failure; (2) patients with severe infections; (3) patients with severe comorbidities; (4) patients with other non-HCC tumors or diseases; (5) patients on medications such as vitamin K or potassium supplements. The study was approved by the Ethics Committee of Sichuan Provincial People’s Hospital (No. 2025.386), and all participants provided informed consent. The study was conducted in accordance with the Declaration of Helsinki (as revised in 2013).
A total of 409 individuals were initially enrolled. After excluding 37 with previously treated HCC, 18 with other malignancies, and 2 with obstructive jaundice, 352 participants remained and were included in the final analysis: 115 HCC, 137 CLD, and 100 HCs.
Measurements of tumor biomarkers
Venous blood (3 mL) was collected from participants into vacuum tubes without anticoagulants, mixed, and centrifuged to separate the serum. AFP and PIVKA-II were initially analyzed using the Abbott Alinity i automated chemiluminescence immunoassay analyzer, employing corresponding reagents. Selected samples were subsequently tested on the Roche cobas e 801 automated chemiluminescence immunoassay analyzer with matching reagents. Undetected serum samples were stored at −80 °C. Abbott AFP levels ≥ 7.00 ng/mL were considered positive, and PIVKA-II levels ≥ 40 mAU/mL were judged positive according to kit instructions. The ASAP score was calculated using the formula: Ln(p/(1-p)) = − 7.5771177 + 0.04666357 × age − 0.57611693 × sex + 0.42243533 × Ln(AFP) + 1.1051891 × Ln(PIVKA-II), where sex = 0 for males and 1 for females. ASAP scores of 0–33.4.4% were classified as low-risk, 33.4%−66.6% as intermediate risk, and 66.7%−100% as high-risk for HCC screening.
For Elecsys AFP, levels ≥ 7.00 ng/mL were considered positive, and the Elecsys PIVKA-II reference range, provided by the reagent manufacturer, defined PIVKA-II ≥ 21.29 ng/mL as positive. The GAAD score was calculated, with scores ≥ 2.57 indicating high risk and scores < 2.57 indicating low risk.
Statistical analysis
Statistical analyses were performed using IBM SPSS Statistics version 26 (IBM Corp., Armonk, NY, USA; https://www.ibm.com/spss) and graphs were generated using GraphPad Prism version 9 (GraphPad Software, San Diego, CA, USA; https://www.graphpad.com).Non-normally distributed measurements were expressed as median (interquartile range) [M(P25, P75)], and count data were expressed as n (%). The Mann-Whitney U test and Kruskal-Wallis H test were used for group comparisons of measurement data, and the chi-square test (χ2) was used for count data. A P-value of < 0.01 was considered statistically significant. Receiver operating characteristic (ROC) curves were plotted, and the area under the curve (AUC), sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were calculated.
Results
1 Baseline characteristics
A total of 409 samples were screened. After exclusions, 352 samples met the criteria and were included in the final analysis: HCC (n = 115), CLD (n = 137), and HCs (n = 100).(Fig. 1).
Fig. 1.
Flowchart of the study. HCC= hepatocellular carcinoma, CLD = chronic liver disease, HCs= Healthy controls.
The median age in the HCC group was 60 years, significantly higher than that in the CLD group (50 years) and the HCs group (45 years) (P < 0.001). A significant gender distribution difference was observed, with both the HCC and CLD groups predominantly male. Chronic hepatitis B was identified as the primary etiology in both the HCC and CLD groups. Among all groups, the median levels of ARCHITECT AFP, ARCHITECT PIVKA-II, ASAP score, Elecsys AFP, Elecsys PIVKA-II, and GAAD scores were significantly higher in the HCC group compared to both the CLD and HCs groups (P < 0.001). The median levels of each biomarker in the CLD group were also significantly higher than those in the HCs group (P < 0.01) (Table 1and Fig. 2).
Table 1.
Baseline Characterization of Study Population by Indicator.
| Characteristic | HCC (n = 115) | CLD (n = 137) | HCs (n = 100) | P value |
|---|---|---|---|---|
| Demographics | ||||
| Age, years | 60(52,67) | 50 (38,58) | 45 (36,51) | < 0.001 |
| Sex, male | 100(87.0) | 104 (75.9) | 74 (74%) | 0.035 |
| Etiology, n(%) | ||||
| HBsAg positive | 89(78.1) | 88(64.2) | — | 0.05 |
| HCV positive | 3(2.6) | 4(2.9) | — | |
| Other | 22(19.3) | 45(32.8) | — | |
| Laboratory values median(IQR) | ||||
| ALT, U/L | 34(22,52) | 33(20.5,57.5) | 23.5(15.25,36.75) | < 0.001 |
| ARCHITECT AFP, ng/mL | 132.22(13.84,1721.43) | 7.8 (3.43,13.89) | 3.19 (2.36,4.46) | < 0.001 |
| ARCHITECT PIVKA-II, mAU/mL | 802.12(127.63,16508) | 39.33 (23.57,69.4) | 25.21 (20.66,29.07) | < 0.001 |
| ASAP %, | 98.33 (86.38,99.97) | 34.68(25.25,61.44) | 18.82 (12.13,25.33) | < 0.001 |
| Elecsys AFP, ng/mL | 104 (13.5,1210) | 7.61 (3.37,13.35) | 3.16 (2.52,4.28) | < 0.001 |
| Elecsys PIVKA-II, ng/mL | 269.1(63.2,2962) | 19.20(15.36,38.55) | 16.45 (14.40,18.60) | < 0.001 |
| GAAD score | 9.68(6.00,9.98) | 1.12(0.52,2.73) | 0.35 (0.14,0.54) | < 0.001 |
| Positivity rate, n (%) | ||||
| ARCHITECT AFP ≥ 7 ng/mL | 91 (79.1) | 83 (60.6) | 4 (4) | < 0.001 |
| ARCHITECT PIVKA-II ≥ 40.24,mAU/mL | 106(92.2) | 65 (47.4) | 1 (1) | < 0.001 |
| ASAP ≥ 33.4%, | 112(97.4) | 74(54) | 7 (7) | < 0.001 |
| ASAP ≥ 66.7%, | 105 (91.3) | 25(18.2) | 0(0) | < 0.001 |
| Elecsys AFP ≥ 7.00ng/mL, | 92 (80.0) | 79 (57.7) | 6 (6) | < 0.001 |
| Elecsys PIVKA-II ≥ 21.29ng/mL | 106 (92.2) | 60(43.8) | 5 (5) | < 0.001 |
| GAAD ≥ 2.57 | 109 (94.8) | 37 (27.0) | 0(0) | < 0.001 |
AFP =alpha-fetoprotein; ALT = alanine aminotransferase; ASAP = Alpha-fetoprotein, Sex, Age, and Protein induced by vitamin K absence-II; GAAD = gender, age, AFP, des-gamma carboxyprothrombin; PIVKA-II = protein induced by vitamin K absence or antagonist-II; HCC= hepatocellular carcinoma; CLD, chronic liver disease; HBV, hepatitis B virus; Other: alcoholic liver disease and non-alcoholic steatohepatitis; HCs= Healthy controls; IQR=interquartile range.
Fig. 2.
Serum concentrations of three serum biomarkers in the overall group: (A) ARCHITECT AFP, ng/mL; (B) Elecsys AFP, ng/mL; (C) RCHITECT PIVKA-II, mAU/mL; (D) Elecsys PIVKA-II, ng/mL; (E) ASAP%; (F) GAAD score, ****p < 0.0001, ***p < 0.001.
AFP was measured according to healthcare industry standards, and PIVKA-II results were interpreted based on the reference ranges specified by each kit. ASAP scores were considered positive at ≥ 33.4% and ≥ 66.7%, while GAAD was considered positive at ≥ 2.57. The positive rates for each indicator were calculated for the HCC, CLD, and HCs groups. As shown in Table 1, the positive rates for all indicators were significantly higher in the HCC group than in both the CLD and HCs groups (P < 0.001). Similarly, in the CLD group, the positive rates for all indicators were significantly higher than those in the HCs group (P < 0.001).
Figure 2E shows the distribution of ASAP scores with two cut-off points (33.4% and 66.7%). Figure 2F presents the distribution of GAAD with a single cut-off point of 2.57, visualizing the patterns across the HCC, CLD, and HCs groups. As depicted in Fig. 2E, at ASAP < 33.4%, HCC risk is low, but CLD and HCC cannot be excluded; at ASAP 33.4%−66.7%, most cases are CLD, with a few HCC and HCs; at ASAP > 66.7%, HCC risk is high, although some CLD cases remain. As presented in Fig. 2F, GAAD < 2.57 indicates low HCC risk but does not exclude it, while GAAD ≥ 2.57 cannot clearly distinguish HCC from some CLD and HCs. In summary, neither ASAP nor GAAD fully excludes HCC in the low-risk range.
Comparison of the diagnostic value of ARCHITECT AFP, ARCHITECT PIVKA-II, ASAP ≥ 33.4%, ASAP ≥ 66.7%, Elecsys AFP, Elecsys PIVKA-II, and GAAD in the diagnosis of HCC
For the group of 352 subjects diagnosed with HCC by radiological imaging or pathology, sensitivity, specificity, and accuracy were calculated using the following positive thresholds: ARCHITECT AFP ≥ 7 ng/mL, ARCHITECT PIVKA-II ≥ 40.24 mAU/mL, ASAP ≥ 33.4%, ASAP ≥ 66.7%, Elecsys AFP ≥ 7.00 ng/mL, Elecsys PIVKA-II ≥ 21.29 ng/mL, and GAAD ≥ 2.57. ASAP ≥ 33.4% demonstrated the highest sensitivity (0.974), but with lower specificity (0.685). ASAP ≥ 66.7% achieved the highest specificity (0.894) and accuracy (0.901). GAAD showed intermediate performance with a sensitivity of 0.947 and specificity of 0.843. Elecsys AFP outperformed ARCHITECT AFP in sensitivity, specificity, and accuracy. Both ARCHITECT PIVKA-II and Elecsys AFP exhibited equal sensitivity, but ARCHITECT PIVKA-II showed higher specificity (Table 2).
Table 2.
Comparison of the sensitivity, specificity and accuracy of various serum biomarkers assays(%).
| Serum biomarkers | Result | HCC | Non-HCC | Sensitivity | Specificity | Accuracy |
|---|---|---|---|---|---|---|
| Cutoff | (n = 115) | (n = 237) | (%) | (%) | (%) | |
| ARCHITECT AFP | Positive (+) | 91 | 87 | 79.1 | 63.2 | 68.4 |
| ≥ 7.00(ng/mL) | Negatives(-) | 24 | 150 | |||
| ARCHITECT PIVKA-II | Positive (+) | 106 | 66 | 92.1 | 72.1 | 78.6 |
| ≥ 40.24(mAU/mL) | Negatives(-) | 9 | 171 | |||
| ASAP | Positive (+) | 112 | 81 | 97.4 | 65.8 | 76.1 |
| ≥ 33.4% | Negatives(-) | 3 | 156 | |||
| ASAP | Positive (+) | 105 | 25 | 91.3 | 89.4 | 90.1 |
| ≥ 66.7% | Negatives(-) | 10 | 212 | |||
| Elecsys AFP | Positive +) | 92 | 85 | 80 | 64,1 | 69.3 |
| ≥ 7.00(ng/mL) | Negatives(-) | 23 | 152 | |||
| Elecsys PIVKA-II | Positive (+) | 106 | 65 | 92.1 | 72.5 | 78.9 |
| ≥ 21.29(ng/mL) | Negatives(-) | 9 | 172 | |||
| GAAD | Positive (+) | 109 | 37 | 94.7 | 84.3 | 87.8 |
| ≥ 2.57 | Negatives(-) | 6 | 200 |
AFP =alpha-fetoprotein; ASAP = Alpha-fetoprotein, Sex, Age, and Protein induced by vitamin K absence-II; GAAD = gender, age, AFP, des-gamma carboxyprothrombin; PIVKA-II = protein induced by vitamin K absence or antagonist-II; HCC= hepatocellular carcinoma; CLD, chronic liver disease; HCs= Healthy controls.
Diagnostic efficacy of different indicators analyzed for diagnostic efficacy in the HCC group versus the non-HCC group
Diagnostic performance for HCC vs. all NON-HCC groups
HCC and non-HCC groups (CLD, HCs) were included in Test Cohort 1 to analyze the diagnostic performance of various indicators. The cut-off values for ARCHITECT AFP, ARCHITECT PIVKA-II, ASAP, Elecsys AFP, Elecsys PIVKA-II, and GAAD were 13.753 ng/mL, 57.925 mAU/mL, 63.93%, 13.45 ng/mL, 27.2 ng/mL, and 2.69, respectively (Table 3). ASAP and GAAD showed similar AUCs: 0.955 (0.934–0.976) vs. 0.958 (0.934–0.976) (P = 0.610) (Fig. 3A). The sensitivity, specificity, PPV, and NPV of the ASAP score were 0.948, 0.882, 0.796, and 0.972, respectively. The corresponding values for GAAD were 0.948, 0.857, 0.762, and 0.971 (Table 3).
Table 3.
Comparison of the diagnostic efficacy of different serum biomarkers in the test cohort.
| Study subgroup | CUT OFF | AUC (95%CI) | Sensitivity(95%CL) | Specificity(95%CL) | PPV(95%CL) | NPV(95%CL) | |||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Overall Group: HCC(n = 115) vs. NON-HCC controls (n = 237) | |||||||||||
| ARCHITECT AFP, ng/mL | 13.735 | 0.847 | (0.798–0.896) | 0.757 | (0.671–0.826) | 0.865 | (0.816–0.903) | 0.731 | (0.645–0.803) | 0.879 | (0.832–0.916) |
| ARCHITECT PIVKA-II, mAU/mL | 57.925 | 0.916 | (0.883–0.950) | 0.887 | (0.816–0.933) | 0.835 | (0.783–0.877) | 0.723 | (0.644–0.791) | 0.938 | (0.897–0.964) |
| ASAP %, | 63.93 | 0.955 | (0.934–0.976) | 0.948 | (0.891–0.976) | 0.882 | (0.834–0.917) | 0.796 | (0.721–0.855) | 0.972 | (0.941–0.987) |
| Elecsys AFP, ng/mL | 13.45 | 0.848 | (0.800–0.897) | 0.757 | (0.671–0.823) | 0.861 | (0.911−0.899) | 0.725 | (0.639–0.797) | 0.879 | (0.831–0.915) |
| Elecsys PIVKA-II, ng/mL | 27.2 | 0.898 | (0.859–0.937) | 0.904 | (0.837–0.946) | 0.818 | (0.765–0.862) | 0.708 | (0.629–0.775) | 0.943 | (0.907–0.970) |
| GAAD score | 2.69 | 0.958 | (0.940–0.976) | 0.948 | (0.891–0.976) | 0.857 | (0.806–0.896) | 0.762 | (0.686–0.825) | 0.971 | (0.939–0.987) |
| Disease Group: HCC(n = 115) vs. CLD controls (n = 137) | |||||||||||
| ARCHITECT AFP, ng/mL | 25.735 | 0.798 | (0.739–0.856) | 0.661 | (0.570–0.741) | 0.883 | (0.819–0.927) | 0.826 | (0.736–0.890) | 0.756 | (0.684–0.816) |
| ARCHITECT PIVKA-II, mAU/mL | 108.14 | 0.876 | (0.832–0.919) | 0.791 | (0.708–0.856) | 0.825 | (0.753–0.879) | 0.791 | (0.708–0.856) | 0.825 | (0.753–0.879) |
| ASAP %, | 73.31 | 0.926 | (0.893–0.958) | 0.904 | (0.837–0.946) | 0.861 | (0.796–0.909) | 0.846 | (0.771–0.899) | 0.915 | (0.854–0.951) |
| Elecsys AFP, ng/mL | 21.9 | 0.801 | (0.744–0.859) | 0.687 | (0.597–0.765) | 0.869 | (0.802–0.915) | 0.814 | (0.726–0.879) | 0.768 | (0.695–0.827) |
| Elecsys PIVKA-II, ng/mL | 33.565 | 0.857 | (0.809–0.905) | 0.861 | (0.786–0.913) | 0.737 | (0.658–0.804) | 0.733 | (0.653–0.801) | 0.863 | (0.789–0.914) |
| GAAD score | 3.295 | 0.928 | (0.899–0.957) | 0.922 | (0.858–0.958) | 0.788 | (0.713–0.848) | 0.785 | (0.709–0.846) | 0.923 | (0.860–0.959) |
| NON-HCC Group: CLD(n = 137) vs.HCs controls (n = 100) | |||||||||||
| ARCHITECT AFP, ng/mL | 6.11 | 0.751 | (0.687–0.815) | 0.635 | (0.522–0.711) | 0.95 | (0.888–0.979) | 0.946 | (0.879–0.977) | 0.655 | (0.575–0.728) |
| ARCHITECT PIVKA-II, mAU/mL | 40.1 | 0.727 | (0.663–0.791) | 0.482 | (0.400–0.565) | 0.99 | (0.946–0.999) | 0.985 | (0.920–0.999) | 0.582 | (0.507–0.654) |
| ASAP %, | 27.65 | 0.833 | (0.784–0.883) | 0.723 | (0.642–0.791) | 0.82 | (0.733–0.883) | 0.846 | (0.770–0.900) | 0.683 | (0.596–0.760) |
| Elecsys AFP, ng/mL | 6.685 | 0.743 | (0.679–0.807) | 0.599 | (0.515–0.677) | 0.94 | (0.875–0.972) | 0.932 | (0.859–0.968) | 0.631 | (0.551–0.704) |
| Elecsys PIVKA-II, ng/mL | 21.25 | 0.675 | (0.607–0.743) | 0.438 | (0.358–0.522) | 0.95 | (0.888–0.979) | 0.923 | (0.832–0.967) | 0.552 | (0.478–0.625) |
| GAAD score | 0.705 | 0.832 | (0.782–0.882) | 0.664 | (0.582–0.738) | 0.88 | (0.802–0.930) | 0.883 | (0.807–0.932) | 0.656 | (0.573–0.732) |
AUC = area under the receiver operating characteristic curve, HCC= hepatocellular carcinoma, GAAD =gender, age, AFP = des-gamma carboxyprothrombin, ASAP=Alpha-fetoprotein, Sex, Age, and Protein induced by vitamin K absence-II; HCs= Healthy controls༛CLD = chronic liver disease༛NON-HCC = CLD and HCs; CI = confidence interval.
Fig. 3.
ROC curves for different serum biomarkers in the test cohort, AUC = area under the receiver operating characteristic curve, HCC= hepatocellular carcinoma, GAAD =gender, age, AFP = des-gamma carboxyprothrombin, ASAP=Alpha-fetoprotein, Sex, Age, and Protein induced by vitamin K absence-II; HCs= Healthy controls༛CLD = chronic liver disease༛NON-HCC = CLD and HCs.
Diagnostic performance for HCC vs. all CLD groups
The HCC and CLD groups were included in Test Cohort 2. In this cohort, the cut-off values for ARCHITECT AFP, ARCHITECT PIVKA-II, ASAP, Elecsys AFP, Elecsys PIVKA-II, and GAAD were 25.735 ng/mL, 108.14 mAU/mL, 73.31%, 21.9 ng/mL, 33.565 ng/mL, and 3.29, respectively (Table 3). The AUCs for ASAP and GAAD were 0.926 (0.893–0.958) vs. 0.928 (0.899–0.957), with no statistically significant difference (P = 0.810) (Fig. 3B). The sensitivity, specificity, PPV, and NPV of the ASAP score were 0.904, 0.861, 0.846, and 0.915, respectively. The corresponding values for GAAD were 0.922, 0.788, 0.785, and 0.923 (Table 3).
Diagnostic performance for CLD vs. all HCs groups
The CLD and HCs groups were included in Test Cohort 3. In this cohort, the cut-off values for ARCHITECT AFP, ARCHITECT PIVKA-II, ASAP, Elecsys AFP, Elecsys PIVKA-II, and GAAD were 6.11 ng/mL, 40.1 mAU/mL, 27.65%, 5.735 ng/mL, 21.25 ng/mL, and 0.705, respectively (Table 3). The AUCs for ASAP and GAAD were 0.833 (0.784–0.883) vs. 0.832 (0.782–0.88), (P = 0.944) (Fig. 3C). The sensitivity, specificity, PPV, and NPV of the ASAP score were 0.723, 0.820, 0.846, and 0.683, respectively. The corresponding values for GAAD were 0.664, 0.880, 0.883, and 0.656 (Table 3).
Follow-up observation
During the study period (September 2024 to November 2025), cases of HCC development within the CLD group were recorded and analyzed. Follow-up data were retrospectively collected from electronic medical records.
During our subsequent data review, we observed that some patients in the CLD group had progressed to HCC : Among the CLD high-risk cohort (ASAP ≥ 66.7%), 25 individuals were identified, with 3 subsequently diagnosed with HCC. Within the medium-risk cohort (33.4%≤ASAP < 66.7%),25 individuals were identified, with 3 were diagnosed with HCC, while none in the low-risk cohort(ASAP < 33.4%)developed HCC. Regarding the GAAD index, the high-risk cohort(GAAD ≥ 2.57) comprised 37 individuals, with 5 diagnosed with HCC, whereas only 1 case of HCC was identified among the 100 individuals in the low-risk cohort(GAAD < 2.57). Of course, this represents only our findings from the past year’s investigations(Table 4).
Table 4.
HCC detection rates by risk stratification in the ASAP and GAAD cohorts.
| Risk Category | ASAP cohort (n = 137) | GAAD cohort (n = 137) | ||
|---|---|---|---|---|
| CLD, n | HCC, n (%) | CLD, n | HCC, n (%) | |
| High risk | 25 | 3 (12.0) | 37 | 5 (13.5) |
| Medium risk | 49 | 3 (6.1) | / | / |
| Low risk | 63 | 0 (0.0) | 100 | 1 (1.0) |
| Total | 137 | 6 (4.4) | 137 | 6 (4.4) |
HCC = hepatocellular carcinoma, GAAD = gender, age, AFP = des-gamma carboxyprothrombin, ASAP = Alpha-fetoprotein, Sex, Age, and Protein induced by vitamin K absence-II; CLD = chronic liver disease.
HCC is one of the deadliest malignant tumors worldwide, ranking as the fifth most common cancer in men and the ninth in women11. Liver cancer mortality shifted from the third highest in 2018 to the second highest in 202012. Many patients present with intermediate or advanced-stage HCC, resulting in a poor prognosis. Epidemiologic data indicate that the annual mortality rate is nearly equal to the annual incidence rate, highlighting the importance of regular HCC screening for high-risk populations9,13. Such screening can enhance early detection, improve diagnostic methods, and increase patient adherence, ultimately leading to higher rates of early diagnosis and better treatment outcomes14. However, the challenge remains in developing effective strategies for early screening and diagnosis to improve the detection rate of early liver cancer in high-risk groups.
While imaging techniques are crucial for diagnosing liver cancer, their sensitivity and specificity in detecting early-stage HCC are limited, heavily influenced by operator experience. Although CT and MRI offer high resolution with contrast-enhanced imaging, they still face challenges in detecting early-stage liver cancer with sufficient sensitivity. These techniques are also costly and require substantial hospital resources, including specialized facilities, equipment, and personnel, making them impractical for large-scale population screening15. Traditionally, AFP has been recommended as a specific tumor marker for HCC screening and early diagnosis, assisting in disease detection and differential diagnosis. However, many cirrhosis patients exhibit persistently elevated AFP levels, sometimes reaching the thousands, without signs of liver cancer for years. Additionally, nearly one-third of advanced HCC patients are AFP-negative16, limiting its diagnostic value. As a result, there has been ongoing interest in identifying other tumor markers associated with HCC to complement AFP and improve early diagnostic performance17.
In recent years, institutions such as the National Medical Products Administration (NMPA) of China, the European Association for the Study of the Liver (EASL), and the American Association for the Study of Liver Diseases (AASLD) have endorsed PIVKA-II as an auxiliary biomarker for HCC diagnosis18–20. PIVKA-II not only complements AFP in liver cancer diagnosis but also shows excellent performance as an individual marker for HCC detection. Combining PIVKA-II with AFP enhances diagnostic efficiency for HCC5,21. Further studies on these biomarkers, combined with artificial intelligence algorithms, can address the predictive limitations of individual markers, thereby improving diagnostic accuracy, risk prediction, and prognosis for cancer patients22.
In recent years, several diagnostic companies have conducted extensive research on digital algorithms based on serum markers for liver cancer. These algorithms, which do not require additional testing, utilize tumor marker results combined with age and gender to create new biomarker models, such as the ASAP and GAAD scores. By incorporating age and gender, these models standardize tumor marker detection results, reflecting a patient’s risk of HCC through a composite score. Digital algorithms demonstrate higher sensitivity and efficacy metrics than multi-marker tests combined with clinical judgment, simplifying disease risk assessment for both clinicians and laboratories8,23.This study aims to compare the ASAP and GAAD digital algorithms and evaluate their performance against AFP and PIVKA-II in two detection systems.
The results for ARCHITECT AFP, ARCHITECT PIVKA-II, ASAP, Elecsys AFP, Elecsys PIVKA-II, and GAAD show that both single tumor markers and algorithm-based scores exhibited significantly higher median levels in the HCC group compared to the CLD group and HCs. Furthermore, levels in the CLD group were also significantly higher than those in HCs. When the ASAP and GAAD models were applied, key diagnostic parameters for HCC—including sensitivity, specificity, and AUC—were consistently higher than those achieved with AFP or PIVKA-II alone, regardless of whether the Abbott or Roche detection platform was used. These findings suggest that, beyond reliance on tumor markers alone, the integration of digital early-detection algorithms is essential for enhancing screening and diagnostic performance, as well as improving risk prediction for cancer patients.
Although both are digital algorithms, ASAP and GAAD differ in cutoff value settings and result interpretation. ASAP uses two cutoff values: 33.4% and 66.7%, while GAAD employs a single cutoff value of 2.57. Findings indicate that when ASAP < 33.4%, the risk is classified as low, with a very low probability of HCC occurrence. When ASAP ≥ 33.4%, sensitivity peaks, but false-positive rates rise significantly. In the 33.4%−66.7% range (moderate risk), the CLD group and a small proportion of the HCC and HCs groups are present. When ASAP ≥ 66.7%, the HCC risk is very high, with peak specificity, although some CLD cases may still be included. When GAAD < 2.57, the risk of HCC occurrence is minimal; however, GAAD ≥ 2.57 similarly fails to distinguish HCC cases from a small proportion of CLD cases. Neither algorithm can completely rule out HCC occurrence in low-risk scenarios.
This study compared the diagnostic performance of the ASAP and GAAD algorithms, revealing no significant difference between them, although each demonstrated strengths in sensitivity and specificity. Beyond diagnostic comparisons, the ASAP algorithm’s classification of negative, gray zone, and positive results diverged considerably from GAAD’s low-risk and high-risk classifications. An ongoing debate centers on whether patients in the gray zone of ASAP classifications should undergo follow-up or enhanced imaging. A subsequent review of trial data spanning over six months revealed that some patients in the CLD group had already been diagnosed with HCC. In the ASAP classification: among high-risk CLD individuals (n = 25), 3 were diagnosed with HCC; in the medium-risk CLD group (n = 49), 3 were diagnosed with HCC; no cases were identified in the low-risk group. In the GAAD classification: among high-risk individuals (n = 37), 5 were diagnosed with HCC, while 1 case was found in the low-risk group (n = 100).
The findings were also compared with previous literature reports. Our laboratory results showed an AFP sensitivity of 0.757 (0.671–0.826) for the ARCHITECT assay and 0.757 (0.671–0.823) for the Elecsys assay. These findings align with the meta-analysis conclusion that nearly 30% of HCC patients consistently test negative for AFP16 The AUC, sensitivity, and specificity of PIVKA-II in diagnosing HCC across various studies have shown some inconsistencies. Previous research reported a sensitivity of 0.768 (0.679–0.829) and specificity of 0.871 (0.815–0.914) for PIVKA-II in HCC diagnosis13. In the present study, the sensitivities of ARCHITECT PIVKA-II and Elecsys PIVKA-II were 0.887 (0.816–0.933) and 0.904 (0.837–0.946), respectively, surpassing values reported in some previous studies. This may be attributed to the inclusion of primarily oncology-referred samples from patients with advanced-stage HCC, where elevated PIVKA-II levels are likely due to abnormalities in vitamin K metabolism-related enzymes induced during hepatocellular malignant transformation5. However, in our final clinical validation cohort 1 for liver cancer diagnostic efficacy, the cutoff value for ASAP% was 63.93%, while the cutoff for GAAD remained below 2.59, both approaching the threshold values of the two algorithms8,24.
Based on our findings, we propose a clinical algorithm for HCC risk stratification using ASAP and GAAD scores. For patients with chronic liver disease:
ASAP score < 33.4% or GAAD score < 2.57 (low risk): continue routine surveillance;
ASAP score 33.4%−66.7% (intermediate risk): shorten surveillance interval;
ASAP score ≥ 66.7% or GAAD score ≥ 2.57 (high risk): immediate contrast-enhanced imaging or comprehensive HCC screening.This strategy may optimize imaging resource allocation and improve early detection rates in high-risk populations.
While this study provides important findings, it also has limitations. As a single-center exploratory study, the study population predominantly consisted of patients from western China. Therefore, our findings may not be directly generalizable to other regions with different etiologies (e.g., NAFLD/NASH in Western countries). External validation in large, multi-center, and diverse cohorts is warranted. We did not stage HCC patients by BCLC. However, the HCC group served as a positive control for the CLD group, as our goal was to validate ASAP and GAAD scores in high-risk populations, not to compare stages.
Although our study focuses on diagnostic screening, combining diagnostic models with prognostic evaluation is an important future direction. As highlighted by Li et al., tumor-associated lymphatic vessel density is a postoperative prognostic biomarker for liver cancer25. Future studies should combine ASAP/GAAD with such biomarkers for better HCC management.
Conclusions
Compared to tumor markers alone, both the ASAP and GAAD digital algorithms demonstrated superior diagnostic performance for hepatocellular carcinoma (HCC) in this study. No significant difference was observed between the ASAP and GAAD models. However, future studies should compare these models in broader prospective cohort studies. Before the clinical implementation of these predictive models, further research needs to assess their accuracy in predicting HCC occurrence in real-world high-risk populations.
Statement of ethics
This research was conducted ethically in accordance with the World Medical Association Declaration of Helsinki.
Author contributions
L-Z and D-YJ designed the study, collected the data, performed laboratory tests, performed the data analysis, and wrote the manuscript . C-BX interpreted the data. D-YJ, C-BX and Y-H are the senior authors of the study and critically revised the manuscript. All the authors approved the final version of the manuscript.
Funding
This word was supported by the Chengdu Science and Technology Program (NO.2024-YF05-01173-SN to C.X).
Data availability
The datasets generated and/or analyzed during the current study are not publicly available due to patient privacy and ethical restrictions (Ethics Committee of Sichuan Provincial People’s Hospital, Approval No. 2025.386). The data are available from the corresponding author upon reasonable request.
Declarations
Competing interests
The authors declare no competing interests.
Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
Footnotes
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Lu Zhang and Chunbao Xie contributed equally to this work.
Contributor Information
Yi Huang, Email: hwuangyi@qq.com.
Dayong Jiang, Email: scsrmyydyj@163.com.
References
- 1.Han, B. et al. Cancer incidence and mortality in China, 2022. J. Natl. Cancer Cent.4, 47–53. 10.1016/j.jncc.2024.01.006 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Yang, J. D. et al. A global view of hepatocellular carcinoma: trends, risk, prevention and management. Nat. Rev. Gastroenterol. Hepatol.16, 589–604. 10.1038/s41575-019-0186-y (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Singh, G. et al. Biomarkers for hepatocellular cancer. World J. Hepatol.12, 558–573. 10.4254/wjh.v12.i9.558 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Forner, A., Reig, M. & Bruix, J. Hepatocellular carcinoma. Lancet391, 1301–1314. 10.1016/S0140-6736(18)30010-2 (2018). [DOI] [PubMed] [Google Scholar]
- 5.Feng, H., Li, B., Li, Z., Wei, Q. & Ren, L. PIVKA-II serves as a potential biomarker that complements AFP for the diagnosis of hepatocellular carcinoma. BMC Cancer. 21, 401. 10.1186/s12885-021-08138-3 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Xie, D. Y. et al. A review of 2022 Chinese clinical guidelines on the management of hepatocellular carcinoma: updates and insights. Hepatobiliary Surg. Nutr.12, 216–228. 10.21037/hbsn-22-469 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Choi, J. et al. Longitudinal assessment of three serum biomarkers to detect very early-stage hepatocellular carcinoma. Hepatology69, 1983–1994. 10.1002/hep.30233 (2019). [DOI] [PubMed] [Google Scholar]
- 8.Yang, T. et al. A novel online calculator based on serum biomarkers to detect hepatocellular carcinoma among patients with hepatitis B. Clin. Chem.65, 1543–1553. 10.1373/clinchem.2019.308965 (2019). [DOI] [PubMed] [Google Scholar]
- 9.Marrero, J. A. et al. Diagnosis, staging, and management of hepatocellular carcinoma: 2018 practice guidance by the American Association for the Study of Liver Diseases. Hepatology68, 723–750. 10.1002/hep.29913 (2018). [DOI] [PubMed] [Google Scholar]
- 10.Kudo, M. et al. Management of hepatocellular carcinoma in Japan: JSH consensus statements and recommendations 2021 update. Liver Cancer. 10, 181–223. 10.1159/000514174 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Hemken, P. M. et al. Validation of a novel model for the early detection of hepatocellular carcinoma. Clin. Proteom.16, 2. 10.1186/s12014-018-9222-4 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Cao, W., Chen, H. D., Yu, Y. W., Li, N. & Chen, W. Q. Changing profiles of cancer burden worldwide and in China: a secondary analysis of the global cancer statistics 2020. Chin. Med. J.134, 783–791. 10.1097/CM9.0000000000001474 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Liu, S. Y. et al. ASAP score versus GALAD score for detection of hepatitis C-related hepatocellular carcinoma: a multicenter case-control analysis. Front. Oncol.12, 1018396. 10.3389/fonc.2022.1018396 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Polpichai, N. et al. Beyond the liver: a comprehensive review of strategies to prevent hepatocellular carcinoma. J. Clin. Med.13, 6770. 10.3390/jcm13226770 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Ayuso, C. et al. Diagnosis and staging of hepatocellular carcinoma (HCC): current guidelines. Eur. J. Radiol.101, 72–81. 10.1016/j.ejrad.2018.01.025 (2018). [DOI] [PubMed] [Google Scholar]
- 16.Harris, P. S. et al. Hepatocellular carcinoma surveillance: an evidence-based approach. World J. Gastroenterol.25, 1550–1559. 10.3748/wjg.v25.i13.1550 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Atiq, O. et al. An assessment of benefits and harms of hepatocellular carcinoma surveillance in patients with cirrhosis. Hepatology65, 1196–1205. 10.1002/hep.28895 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Department of Medical Administration of National Health Commission. Guidelines for diagnosis and treatment of primary liver cancer (2024). Chin. J. Hepat. Surg. (Electron Ed). 13, 407–449. 10.3877/cma.j.issn.2095-3232.2024.04.001 (2024). (In Chinese). [Google Scholar]
- 19.European Association for the Study of the Liver. EASL clinical practice guidelines: management of hepatocellular carcinoma. J. Hepatol.69, 182–236. 10.1016/j.jhep.2018.03.019 (2018). [DOI] [PubMed] [Google Scholar]
- 20.Heimbach, J. K. et al. AASLD guidelines for the treatment of hepatocellular carcinoma. Hepatology67, 358–380. 10.1002/hep.29086 (2018). [DOI] [PubMed] [Google Scholar]
- 21.Ren, T. et al. Validation of combined AFP, AFP-L3, and PIVKA II for diagnosis and monitoring of hepatocellular carcinoma in Chinese patients. Heliyon9, e21906. 10.1016/j.heliyon.2023.e21906 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Calderaro, J., Seraphin, T. P., Luedde, T. & Simon, T. G. Artificial intelligence for the prevention and clinical management of hepatocellular carcinoma. J. Hepatol.76, 1348–1361. 10.1016/j.jhep.2022.01.014 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Piratvisuth, T. et al. Development, clinical validation and implementation of a novel algorithmic score (GAAD) for the detection of early-stage hepatocellular carcinoma [abstract]. J. Hepatol.78, S507–S508. 10.1016/S0168-8278(23)01252-7 (2023). [Google Scholar]
- 24.Piratvisuth, T. et al. Development and clinical validation of a novel algorithmic score (GAAD) for detecting HCC in prospective cohort studies. Hepatol. Commun.7, e0317. 10.1097/HC9.0000000000000317 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Li, J. et al. Tumor-associated lymphatic vessel density is a postoperative prognostic biomarker of hepatobiliary cancers: a systematic review and meta-analysis. Front. Immunol.15, 1519999 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
The datasets generated and/or analyzed during the current study are not publicly available due to patient privacy and ethical restrictions (Ethics Committee of Sichuan Provincial People’s Hospital, Approval No. 2025.386). The data are available from the corresponding author upon reasonable request.