Assay performance verification and reference interval establishment for PIVKA-II on the Roche cobas e 801

To validate the analytical performance of the Roche cobas e 801 chemiluminescent immunoassay for Protein induced by vitamin K absence or antagonist-II (PIVKA-II), establish age-/gender-specific reference intervals for healthy adults in the Gansu population, and evaluate its concordance with the Abbott Alinity i system currently used in our laboratory.

Precision (intra-run/inter-run coefficients of variation [CVs]), linearity (0.50-12000 ng/mL), and clinical reportable range (via 10-fold dilution with 80%-120% recovery criteria) were verified following Clinical and Laboratory Standards Institute (CLSI) and Chinese health industry standards. Agreement between the cobas e 801 and Alinity i systems was assessed using Passing-Bablok regression and Bland-Altman analysis (n = 173). Reference intervals were established based on 2253 healthy participants (≥ 18 years, permanent residents of Lanzhou). The 2.5th-97.5th percentiles were calculated using a nonparametric method, and decision tree analysis was used to determine age stratification cutoffs.

Intra-run coefficients of variation ranged from 1.36% to 1.57%, and inter-run CVs from 1.13% to 1.16%, all below CLIA'88 total allowable error and manufacturer claims (except for intra-run high-value quality control (QC), which was slightly higher than the manufacturer's specification). The linear regression equation was Y = 0.99X + 6.34 (R²= 0.99), and deviations at all concentration points were < 1/2 Total allowable error (Tea) (11%), meeting linearity requirements. Recoveries after 10-fold dilution ranged from 89.62% to 92.16%, validating the clinical reportable range of 15.30-91533.33 ng/mL. Verification using 20 healthy individuals showed that 100% of results were below the manufacturer's reference interval upper limit (< 28.4 ng/mL). Method comparison revealed that the intercept 95% confidence interval (CI) (-2.16 to 3.42) included 0, indicating no significant constant (systematic) error, while the slope 95% CI (0.4693 to 0.6925) did not include 1, indicating significant proportional bias. Bland-Altman analysis showed that 96.50% (167/173) of data points fell within the limits of agreement, indicating good agreement between the two methods. The reference interval for males (≥ 18 years) was 9.63-29.10 ng/mL, with no age stratification required. For females, age-stratified reference intervals were established: 9.50-26.30 ng/mL for ages 18-54 years, and 9.42-30.00 ng/mL for ages ≥ 55 years. We recommend a unified reference interval of 9.52-29.10 ng/mL for routine clinical use. Age- and sex-stratified intervals are statistically significant but have modest clinical impact (∼3 ng/mL difference in upper limits).

The PIVKA-II assay on the Roche cobas e 801 system meets laboratory quality requirements for analytical performance and is suitable for clinical application. Due to significant proportional bias compared with the Abbott Alinity i system, system-specific reference intervals are recommended, and the two systems should not be used interchangeably. For results approaching the upper reference limit, clinical follow-up is advised. The established local reference intervals can provide more targeted decision-making support for hepatocellular carcinoma screening.