Validation Study of Domestic Protein Induced by Vitamin K Antagonist-II Diagnostic Reagents: A Multicenter Prospective Clinical Trial

Li Xiaohua¹, Men Kun², Tian Xuezhi³, Zhang Baoping⁴, Cao Yang²

¹Clinical Trial Institution, the Second Hospital of Tianjin Medical University, Tianjin 300211, China; ²Clinical Laboratory, the Second Hospital of Tianjin Medical University, Tianjin 300211, China; ³Clinical Laboratory, Shanxi Cancer Hospital, Taiyuan 030013, China; ⁴Clinical Laboratory, the Affiliated Hospital of Inner Mongolia Medical University, Hohhot 010030, China

Corresponding author: Cao Yang, Email: ttykcaochen@126.com
DOI: 10.3760/cma.j.cn102012-20250326-00016
Received: 2025-03-26
Edited by: Xia Shuang

Abstract

Objective To verify the detection consistency between domestic PIVKA-II diagnostic reagents and the original product.

Methods A prospective clinical trial was conducted across three hospitals. From November 15, 2021, to April 24, 2024, a total of 1,329 blood samples were collected from the laboratories of three hospitals. According to inclusion and exclusion criteria, 1,278 samples met the trial protocol requirements, with 385, 334, and 559 samples included from the three laboratories, respectively. After assigning serial numbers, samples were tested using both domestic PIVKA-II diagnostic reagents (test reagents) and original PIVKA-II diagnostic reagents (control reagents). Following the "Guidelines for Clinical Trial Techniques of In Vitro Diagnostic Reagents," regression coefficients, intercepts, and correlation coefficients were used as clinical evaluation indices. Wilcoxon test, linear regression, Passing-Bablok regression, and Bland-Altman analysis were employed to determine testing consistency between the two products. Spearman correlation analysis was applied, and univariate and multivariate analyses were conducted to identify factors influencing the differences.

Results All 1,278 protocol-compliant samples were included in statistical analysis, including 483 liver tumor samples. The Wilcoxon test revealed no significant differences between test and control reagents in either the entire population or the liver tumor population (entire population: Z = -1.414, P = 0.157; liver tumor: Z = -1.238, P = 0.216). The correlation coefficient between the two reagents in the entire population was 0.9985. Detection values showed statistically significant differences across gender, laboratory, and disease diagnosis (all P < 0.05). Bland-Altman analysis demonstrated good symmetry in both absolute and relative bias plots, with no more than 5% of samples exceeding the 95% consistency limits. Linear regression and Passing-Bablok regression indicated good correlation between the two reagents. Using control reagent results as the independent variable and test reagent results as the dependent variable, the linear regression equation was y = 1.0007x + 0.215 (R² = 0.9999, P < 0.05), while the Passing-Bablok regression equation was y = 1.000x + 0.040 (R² = 0.9999, P < 0.05). The detection differences between the two reagents exhibited a peaked distribution. After logarithmic transformation, univariate analysis showed that gender, disease diagnosis, and laboratory factors affected the differences. Logarithmic transformation and generalized linear regression analysis revealed that disease diagnosis and detection values had statistically significant effects on the differences, while interfering samples did not.

Conclusions The consistency between the two diagnostic reagents is good, and PIVKA-II detection values and differences are primarily influenced by patients' disease status.

Keywords: Liver neoplasms; Reagent kits, diagnostic; Protein induced by vitamin K antagonist-II; In vitro diagnostic reagent; Clinical trial

Funding: Tianjin Science and Technology Plan Project (18JCZDJC36000)

Serum abnormal prothrombin, also known as protein induced by vitamin K absence or antagonist-II (PIVKA-II), is an important indicator for liver tumor diagnosis that complements alpha-fetoprotein. PIVKA-II primarily exists in the serum of liver cancer patients and represents a precursor protein of human prothrombin that is incompletely carboxylated in the liver. When carboxylase inhibitors appear in hepatocytes, patients' serum PIVKA-II levels generally increase, suggesting malignant transformation of liver cells. PIVKA-II detection can effectively predict the risk of liver tumor development. The "Guidelines for Diagnosis and Treatment of Primary Liver Cancer (2024 Edition)" recommends PIVKA-II as an early diagnostic marker for liver cancer, particularly in populations with negative serum alpha-fetoprotein. Research also indicates that PIVKA-II has value in evaluating short-term efficacy after transcatheter arterial chemoembolization (TACE) in liver tumor patients.

The diagnostic reagent kit for this indicator was developed and marketed by Abbott GmbH & Co. KG. Due to its high price and special instrument requirements, its普及性 remains insufficient. Developing domestic products of the same type aligns with the research and development concept of "unmet clinical needs." The original product's instructions recommend establishing normal value ranges for each laboratory based on sex, age, and ethnicity. However, literature presents conflicting views on the extent to which biological characteristics affect this indicator's levels. Therefore, we conducted a consistency comparison between the original product (control reagent) and the domestic reagent (test reagent) and analyzed factors influencing detection results to provide a basis for rational application in clinical testing.

Materials and Methods

Study Design

This study was a diagnostic clinical trial evaluating the consistency between test reagent results and control diagnostic reagent results from samples collected at a single time point. The collected blood samples were residual specimens from routine clinical testing. While test reagents were used for sample detection, subjects also underwent routine clinical diagnosis and laboratory testing. Test reagent results were not used for patient management and did not affect clinical decision-making. The trial was conducted at three hospitals in China—two general hospitals and one specialized cancer hospital. The medical laboratory professionals and principal investigators at all three hospitals completed registration in the medical device clinical trial institution filing management information system of the National Medical Products Administration before the trial began. The clinical trial protocol was written according to the "Guidelines for Clinical Trial Techniques of In Vitro Diagnostic Reagents" and submitted to local medical products administrations for registration before trial initiation, with no updates or revisions during the trial (filing number: 20210030).

Ethics Review

In accordance with the "Regulations of the People's Republic of China on the Management of Human Genetic Resources," the trial project was approved by the Office of Human Genetic Resources Management of China before initiation (approval number: Guoke Yiban Shenzi [2021] CJ2266). This study was approved by the Medical Ethics Committee of the Second Hospital of Tianjin Medical University (approval number: 2021K138), the Ethics Committee for Drug and Medical Device Clinical Trials of Shanxi Cancer Hospital (approval number: QX2021005), and the Ethics Committee of the Affiliated Hospital of Inner Mongolia Medical University (approval number: SY2021042), following the 2013 Declaration of Helsinki. The trial began in 2021. According to Article 39 of the "Ethical Review Measures for Biomedical Research Involving Humans (2016 Edition)"—"For research using identifiable human materials or data where the subject can no longer be located, and the research project does not involve personal privacy or commercial interests"—this project was approved by the ethics committees for exemption from informed consent.

Quality Control

Currently, the National Health Commission's Clinical Testing Center does not organize inter-laboratory quality assessment for serum PIVKA-II. However, all three laboratories participate in and pass inter-laboratory quality assessments for other tumor marker projects organized by the National Health Commission's Clinical Testing Center. Both test and control reagents used their respective kit-matched quality control products for quality control and passed on each testing day. All three laboratories followed a unified trial protocol and used uniform reagents and instruments. Participating researchers were authorized by their respective laboratory principal investigators and received training on the protocol and sample detection standard operating procedures before trial initiation. Blood samples were tested regularly according to sample collection and storage periods, with each batch completing calibration, quality control, and detection processes.

Sample Size Calculation

The expected positive rate (π) for this project was 30%, with estimated error (E) generally not exceeding 10% (5% was used for this trial). At 95% confidence interval (CI), the minimum sample size n for each center was calculated using formula (1): n = 1.962 × 0.3(1-0.3)/0.05² = 322.69. Considering a 5% sample exclusion rate, the minimum sample size for each center in this clinical trial was set at 340 cases, with a minimum total sample size of 1,020 across three clinical trial centers.

Sample Collection

From November 15, 2021, to April 24, 2024, residual serum samples from routine clinical testing or frozen samples for clinical testing were collected at the Second Hospital of Tianjin Medical University, Affiliated Hospital of Inner Mongolia Medical University, and Shanxi Cancer Hospital. Samples were screened based on clinical diagnostic information and potential departments, with eligible samples selected by entering data into case report forms.

Diagnostic Information and Related Sample Requirements: (1) PIVKA-II test results positive, i.e., ≥300 samples outside the normal reference range; (2) ≥540 samples from patients with liver-related diseases (such as hepatitis B, liver cirrhosis) and other diseases, including renal insufficiency, hyperlipidemia, osteoporosis, rheumatoid arthritis, respiratory failure, anemia, arrhythmia, pneumonia, intestinal obstruction, cytopenia, lupus erythematosus, cerebral infarction, etc.; (3) ≥40 samples with potential specific interference (hemolysis, lipemia, jaundice, rheumatoid factor, or other types); (4) ≥60 samples positive for potentially interfering tumor markers (including alpha-fetoprotein, CA125, CA153, CA19-9, or other types) from tumors in the stomach, endometrium, rectum, esophagus, lung, ovary, breast, nasopharynx, bladder, cervix, prostate, and other sites.

Inclusion Criteria: (1) Complete sample information, including sample number, sex, age, and clinical diagnosis; (2) Each serum sample volume not less than 0.5 ml; (3) Samples from inpatients, outpatients, and physical examination personnel, covering the detection range of this reagent and including high, medium, and low value samples (no proportion requirement,预判 based on diagnostic information).

Exclusion Criteria: (1) Contaminated samples; (2) Samples collected at different times from the same patient (only the first collected sample was retained, with duplicate samples excluded).

Exclusion Criteria During Trial: (1) Samples with insufficient volume for detection due to operational errors; (2) Samples that could not complete the entire testing process due to other factors; (3) Samples missing any required information from original clinical research records before statistical analysis; (4) Samples deemed necessary to exclude by researchers (such as those exceeding the detection upper limit).

Sample Storage and Coding

Collected residual samples were serum type. Samples were stable for 24 hours at 10–30°C, 5 days at 2–8°C, and 6 months below -20°C, with no more than 2 freeze-thaw cycles and thorough mixing after thawing.

Collected samples were coded by researchers not involved in detection. Each sample was assigned a unique serial number and tested using both test and control reagents. Laboratory personnel responsible for detection were blinded to sample diagnostic information.

Diagnostic Reagents

The test reagent was the PIVKA-II diagnostic reagent kit (magnetic particle chemiluminescence method) produced by Tianjin Huaketai Biotechnology Co., Ltd., with reagent batch number 20220415 and linear range of 20.00–20,000.00 mAU/mL. The matched calibrator (batch number 20220411) and quality control product (batch number 20220411) were produced by the same company. The control reagent was the PIVKA-II assay kit (chemiluminescent microparticle immunoassay) produced by Abbott GmbH & Co. KG, with reagent batch number 37941LP36 and linear range of 0.00–30,000.00 mAU/mL, covering the test reagent's linear range. The matched calibrator (batch number 28945LP23) and quality control product (batch number 28947LP23) were produced by the same company. The control reagent has good sensitivity and specificity, with linear detection range and sensitivity similar to the test product. The normal reference value for both test and control reagents was <40 mAU/mL.

Detection Instruments

The test reagent was matched with the fully automatic chemiluminescence immunoanalyzer (model: Shine i2100) produced by Shenzhen Yingkai Biotechnology Co., Ltd. The control reagent was matched with the analyzer (model: Alinity i) produced by Abbott GmbH & Co. KG.

Statistical Methods

Data were analyzed using R 4.4.2 statistical software. Normally distributed quantitative data were expressed as mean ± standard deviation (s x ±), with independent samples t-test or ANOVA for inter-group comparisons. Non-normally distributed quantitative data were expressed as median (P25, P75), with Mann-Whitney U test or Kruskal-Wallis H test for inter-group comparisons. Categorical data were expressed as frequency and percentage. Comparison of detection values between test and control reagents used paired t-test or Wilcoxon test based on normal distribution. According to the "Guidelines for Clinical Trial Techniques of In Vitro Diagnostic Reagents," correlation analysis of quantitative detection results used Pearson or Spearman correlation based on data distribution normality. Bland-Altman method was used to calculate consistency limits and evaluate consistency between the two detection results. Linear regression and Passing-Bablok regression were performed using control reagent results as the independent variable and test reagent results as the dependent variable to evaluate consistency between the two methods. Outliers were assessed by summing absolute differences between test and control reagent results and calculating the average. For the i-th sample, if the absolute difference exceeded the outlier detection limit of 4, it was considered an outlier. Similarly, the average of relative differences was calculated, with a relative outlier detection limit of 4. Any sample exceeding both absolute and relative difference limits was determined to be an outlier. Univariate and multivariate analyses were used to identify factors influencing detection differences. Two-sided tests were used, with P < 0.05 considered statistically significant.

Results

Basic Sample Information

A total of 1,329 samples were obtained during the study period. Among them, 48 samples exceeded the detection range, 3 lacked clinical diagnosis, and 1,278 samples met inclusion criteria. Sample and corresponding patient characteristics are shown in Table 1 [TABLE:1]. Gender and diagnosis differences among patients at the three trial centers were statistically significant (all P < 0.001). PIVKA-II detection concentrations did not follow a normal distribution (Kolmogorov-Smirnov test: D = 0.406, P < 0.001; Shapiro-Wilk test: W = 0.368, P < 0.001).

Comparison of Two Reagents

Since PIVKA-II detection results did not follow a normal distribution, Spearman correlation analysis was applied. The overall correlation coefficient between test and control reagent results was 0.9985. Correlations between the two reagents within each laboratory were also good. Paired Wilcoxon tests showed no statistically significant differences between test and control reagents, with no significant differences within each laboratory either (Table 2 [TABLE:2]).

Outlier Analysis Results

The average sum of absolute differences between the two reagents was 9.749, with an outlier detection limit of 38.994; 96 samples exceeded the absolute difference detection limit. The average sum of relative differences was 0.018, with a relative outlier detection limit of 0.071; 21 samples exceeded the relative difference detection limit. However, no sample simultaneously exceeded both absolute and relative difference detection limits.

Bland-Altman Analysis Results

Both absolute and relative bias plots showed good symmetry (Figure 1 [FIGURE:1]). The absolute difference between the two diagnostic reagents was 0.0400 (-0.5100, 0.5800) mAU/mL, with 95% consistency limits of (-61.212, 63.018) mAU/mL. 4.38% of samples fell outside the 95% consistency limits. The relative difference between the two diagnostic reagents was 0.0013 (-0.0121, 0.0138), with 95% consistency limits of (-0.056, 0.060). 1.88% of samples fell outside the 95% consistency limits.

Regression Analysis Results

Scatter plots of test and control reagent detection results were drawn (Figure 2 [FIGURE:2]). Using control reagent results as the independent variable and test reagent results as the dependent variable, the linear regression equation was y = 1.0007x + 0.215 (R² = 0.9999, P < 0.05). The Passing-Bablok regression equation was y = 1.000x + 0.040 (R² = 0.9999, P < 0.05), indicating a positive correlation between the two reagents' test results with good overall and laboratory-specific correlations.

Normality of Absolute and Relative Differences

Absolute differences in reagent detection did not follow a normal distribution (Kolmogorov-Smirnov test: D = 0.366, P < 0.001; Shapiro-Wilk test: W = 0.437, P < 0.001). The absolute difference histogram showed a peaked distribution, and the Q-Q plot was S-shaped (Figure 3 [FIGURE:3]). Relative differences also did not follow a normal distribution (Kolmogorov-Smirnov test: D = 0.106, P < 0.001; Shapiro-Wilk test: W = 0.699, P < 0.001). After logarithmic transformation, the normality of absolute differences improved in both histogram and Q-Q plot. Absolute differences were logarithmically correlated with test results: y = 0.3965ln(x) - 1.7991 (R² = 0.7288).

Analysis of Factors Related to Detection Differences

Univariate analysis showed that gender, laboratory, and disease diagnosis were not correlated with absolute and relative differences between the two reagents (all P > 0.05) (Table 3 [TABLE:3], Figure 4 [FIGURE:4]). After logarithmic transformation of absolute differences, they became associated with gender, laboratory, and disease diagnosis (all P < 0.05). Interference factors were not correlated with detection results (all P > 0.05) (Table 4 [TABLE:4]). The selected samples covered all age groups (Table 5 [TABLE:5]). A high-density scatter plot of age versus absolute differences was drawn (Figure 5 [FIGURE:5]). Spearman correlation analysis showed correlation coefficients of R = 0.001 (P = 0.967) between age and test reagent, R = -0.001 (P = 0.995) between age and control reagent, and R = -0.0241 (P = 0.389) between age and absolute differences between the two reagents.

To reduce multiplicity and collinearity, multivariate analysis only examined absolute differences. Both logarithmic transformation and generalized linear regression with Gamma distribution were used. Based on univariate analysis results and findings from other studies, sex, age, disease diagnosis, PIVKA-II detection value, and interference sample status were included in multivariate analysis. Inter-laboratory differences, being within the scope of inter-laboratory quality assessment, were not included as multivariate analysis variables. Multivariate analysis for each laboratory showed that PIVKA-II detection value and disease diagnosis had statistically significant effects on differences (all P < 0.05) (Table 6 [TABLE:6]).

Discussion

As a liver tumor marker, PIVKA-II diagnostic reagents must have good sensitivity and specificity, and corresponding clinical trials should be conducted before registration and marketing. This study compared consistency between domestic and original reagents following the "Guidelines for Clinical Trial Techniques of In Vitro Diagnostic Reagents," conducting correlation analysis, paired tests, outlier analysis, Bland-Altman analysis, and regression analysis. Since the difference distribution between the two diagnostic reagents showed a peaked distribution and did not follow a normal distribution, not fully meeting Bland-Altman analysis conditions, we nevertheless used the recommended Bland-Altman analysis from the guidelines given good distribution symmetry, while also conducting segmented statistics. Since PIVKA-II detection values showed skewed distribution, both linear regression and Passing-Bablok regression were performed. Compared with linear regression, Passing-Bablok regression does not require normally distributed data and is more suitable for comparing consistency between two reagents. This study was conducted simultaneously in three different types of medical institution laboratories, including both general and specialized cancer hospitals, with collected samples possessing certain diversity. Therefore, during consistency analysis, we examined not only overall consistency between the two diagnostic reagents but also intra-laboratory consistency, with correlation and regression analysis results indicating good consistency between the two diagnostic reagents.

This study also examined whether differences between the two diagnostic reagents were related to age, sex, and laboratory factors. Consistent with literature reports on apparently healthy populations, PIVKA-II detection results differed between sexes and disease conditions, with sex distribution varying across diseases; however, no significant correlation was found between age and detection results. Univariate non-parametric analysis showed these factors were unrelated to differences. However, considering that the distribution of detection differences did not meet normality assumptions, non-parametric analysis methods were not very sensitive. Therefore, data were logarithmically transformed, after which sex and disease condition were unrelated to detection differences. Multivariate analysis results showed that after adjusting for detection value effects, sex was unrelated to detection differences between the two reagents. In this study, different laboratories enrolled patients with different disease distributions, which may be the main reason for laboratory differences observed in univariate analysis.

From an economic perspective, the winning bid price for the control reagent's matching instrument is approximately 600,000 RMB, while the test reagent's matching instrument has obtained registration approval but has no public quotation, with similar domestic products priced at approximately 200,000 RMB. Scholars have investigated prices of such domestic and imported reagents, finding domestic reagents overall cheaper than imported ones. Although the test reagent used in this study has not yet been priced, the overall trend shows price advantages for domestic reagents and instruments.

Due to existing literature, this study did not involve analysis of differences in normal value ranges for apparently healthy populations, focusing more on the target indication, related diseases, and interference populations. Only Chinese population blood samples were included, so the relationship between ethnicity and detection differences could not be evaluated. This study's protocol was designed according to guideline principles, examining consistency between domestic and original reagents, with selected samples covering relevant disease subgroups. However, age and sex subgroups were not pre-specified, and exploratory multivariate analysis was used during the data analysis phase to evaluate factors related to differences. All three participating centers followed the same trial protocol, but all tested blood samples were from their own centers without cross-testing. Whether laboratory factors affect assessment of detection differences still requires verification through inter-laboratory quality assessment.

Conflict of Interest

None of the authors hold stocks in Tianjin Huaketai Biotechnology Co., Ltd., nor do they serve as consultants or provide consulting/training services to the company. Tianjin Huaketai Biotechnology Co., Ltd. was responsible for providing reagents, instruments, and trial funding during the trial process, without influencing trial implementation or data analysis.

Acknowledgments

The clinical trial institution research teams from the Second Hospital of Tianjin Medical University, Affiliated Hospital of Inner Mongolia Medical University, and Shanxi Cancer Hospital participated in this clinical trial and provided data. Tianjin Huaketai Biotechnology Co., Ltd. provided reagents and other trial materials.

Author Contributions

Li Xiaohua: Study design, statistical analysis, manuscript writing; Men Kun: Trial implementation, data collection; Tian Xuezhi, Zhang Baoping: Trial implementation, data collection; Cao Yang: Study design, protocol review, manuscript revision.

References

[1] Wang Guiqiang, Wang Fusheng, Zhuang Hui, et al. Guidelines for the prevention and treatment of chronic hepatitis B (2019 edition)[J]. Chinese Journal of Viral Diseases, 2020, 10(1): 1-25. DOI: 10.16505/j.2095-0136.2019.0097.

[2] Feng H, Li B, Li Z, et al. PIVKA-II serves as a potential biomarker that complements AFP for the diagnosis of hepatocellular carcinoma[J]. BMC Cancer, 2021, 21(1): 401. DOI: 10.1186/s12885-021-08138-3.

[3] Ng WY, Lim D, Tan SY, et al. The role of PIVKA-II in hepatocellular carcinoma surveillance in an Asian population[J]. Ann Acad Med Singap, 2023, 52(2): 108-110. DOI: 10.47102/annals-acadmedsg.2022351.

[4] Yu L, Wang X, Zhang X, et al. Construction and evaluation of a nomogram model for predicting hepatocellular carcinoma incidence based on serum abnormal prothrombin and alpha-fetoprotein[J]. Chinese Journal of Hepatobiliary Surgery, 2025, 31(1): 1-5. DOI: 10.3760/cma.j.cn113884-20240530-00165.

[5] Luo H, Ren W. Predictive value of CT signs combined with serum abnormal prothrombin for solitary liver cancer microvascular invasion[J]. Journal of Xuzhou Medical University, 2025, 45(2): 151-156. DOI: 10.12467/j.issn.2096-3882.20230960.

[6] Medical Administration Department, National Health Commission of the People's Republic of China. Guidelines for diagnosis and treatment of primary liver cancer (2024 edition)[J]. Chinese Journal of Hepatology, 2024, 32(7): 581-630. DOI: 10.3760/cma.j.cn501113-.

[7] Zha J, Guo J. Value of abnormal prothrombin in evaluating short-term efficacy after transcatheter arterial chemoembolization for hepatocellular carcinoma[J]. Journal of Clinical Hepatology, 2020, 36(9): 1990-1993. DOI: 10.3969/j.issn.1001-5256.2020.09.016.

[8] Yan C, Hu J, Yang J, et al. Serum ARCHITECT PIVKA-II reference interval in healthy Chinese adults: sub-analysis from a prospective multicenter study[J]. Clin Biochem, 2018, 54: 32-36. DOI: 10.1016/j.clinbiochem.2018.02.007.

[9] Jia J. Statistics. 6th ed[M]. Beijing: China Renmin University Press, 2016: 176.

[10] National Medical Products Administration. Notice on the release of technical guidelines for clinical trials of in vitro diagnostic reagents (No. 72 of 2021)[EB/OL]. (2021-09-16) [2025-03-20]. https://www.nmpa.gov.cn/ylqx/ylqxggtg/20210927152837140.html.

[11] Beijing Medical Products Administration. Guidelines for clinical trials of Class II in vitro diagnostic reagents in Beijing (2016)[EB/OL]. (2015-12-19) [2025-03-20]. https://yjj.beijing.gov.cn/yjj/resource/cms/2016/07/2016071818013682054.doc.

[12] National Health and Family Planning Commission. Ethical review measures for biomedical research involving humans[EB/OL]. (2016-10-12) [2025-03-20]. https://www.gov.cn/gongbao/content/2017/content_5227817.htm.

[13] Yang T, Xing H, Wang G, et al. A novel online calculator based on serum biomarkers to detect hepatocellular carcinoma among patients with hepatitis B[J]. Clin Chem, 2019, 65(12): 1543-1553. DOI: 10.1373/clinchem.2019.308965.

[14] Passing H, Bablok. A new biometrical procedure for testing the equality of measurements from two different analytical methods. Application of linear regression procedures for method comparison studies in clinical chemistry, Part I[J]. Chem Clin Biochem, 1983, 21(11): 709-720. DOI: 10.1515/cclm.1983.21.11.709.

[15] Wei B, Zheng Y, Li L, et al. Establishment of sex-specific reference intervals for PIVKA-II in Southwest China: a real-world data analysis[J]. Ann Clin Biochem, 2025, 62(3): 202-207. DOI: 10.1177/00045632241306074.

[16] Heilongjiang Public Resources Trading Network. Results announcement for fully automated chemiluminescence immunoassay system (fully automated immunoassay analyzer + fully automated sample processing system), fully automated biochemical analyzer, and fully automated specific protein analyzer for the First Affiliated Hospital of Harbin Medical University[EB/OL]. (2024-07-01) [2025-04-29]. https://ggzyjyw.hlj.gov.cn/jyfwdt/003002/003002002/003002002004/20240701/2c9086669068efde01906cccf91541d1.html.

[17] China Government Procurement Network. Tender announcement for fully automated chemiluminescence immunoanalyzer (second bidding)[EB/OL]. (2024-08-19) [2025-04-29]. http://www.ccgp.gov.cn/cggg/dfgg/gkzb/202408/t20240819_22931360.htm.

[18] Chen Y, Chen D. Impact of Sichuan centralized online sunshine procurement policy on hospital in vitro diagnostic reagent procurement prices[J]. China Pharmaceuticals, 2018, 27(20): 80-83.