国产异常凝血酶原诊断试剂验证研究：多中心前瞻性临床试验

Objective To verify the consistency between domestically produced PIVKA- diagnostic reagents and the original product.Methods A prospective clinical trial was conducted in three hospitals and 1 329 blood samples were collected from the laboratories from November 15, 2021 to April 24, 2024. According to inclusion and exclusion criteria, 1 278 samples met the requirements of the trial protocol. Three laboratories collected 385, 334, and 559 samples, respectively. After the sample serial number was assigned, test reagents (short for domestically produced PIVKA- diagnostic reagents) and control reagents (short for original PIVKA- diagnostic reagents) were used for detection serum sample. According to Guideline of Clinical Trial Techniques for In Vitro Diagnostic Reagents, regression coefficient, intercept, and correlation coefficient were used as clinical evaluation index. Wilcoxon testlinear regression, Passing-Bablok regression, and Bland-Altman analysis were used to determine the detection consistency between two reagents. Spearman analysis was used to determine the correlation of PIVKA- concentration detected by the two reagents. Univariate analysis and multivariate analysis were used to determine the factors influencing differences between two diagnostic reagents.Results Totally 1 278 samples that met the protocol were included in statistical analysis, including 483 liver tumor samples. In both the entire population and the liver tumor population, the Wilcoxon test revealed that there were no statistical differences between test and control reagents for PIVKA- concentration detection (entire population: Z=1.414, P=0.157; liver tumor population: Z=1.238, P=0.216). The Spearman correlation coefficient between the test reagent and the control reagent in the entire population was 0.998 5. There were statistically significant differences in detection values among genders, laboratories, and disease diagnoses (all P<0.05). The Bland-Altman analysis demonstrated good symmetry in both absolute and relative bias plots, and no more than 5% of the samples exceeded the 95% consistency limit. Linear regression and Passing-Bablok regression indicated a good correlation between the two reagents. The linear regression analysis using the control reagent test results as the independent variable and the test reagent results as the dependent variable yielded the fitted equation: y=1.0007x+0.215 (R2=0.999 9, P<0.05), while the Passing-Bablok regression fitting equation was y=1.000x+0.040 (R2=0.999 9, P<0.05). The detection differences between the two reagents exhibited a peak distribution; thus, differences didnt follow a normal distribution. After logarithmic transformation, univariate analysis showed that gender, disease diagnosis, and laboratory factors influenced the differences. Logarithmic transformation and generalized linear regression analysis indicated that disease diagnosis and PIVKA- detection values had a statistically significant impact on differences, while interfering samples had no statistically significant impact on differences.Conclusions The consistency between the two reagents is good, and the detection value and difference of PIVKA- are mainly affected by the patient's disease status.