Establishment of PIVKA-฀ Reference Intervals From Hospital-stored Data: A Comparison Analysis

Objective The authors aimed to explore methods to establish indirect reference 46 intervals for PIVKA - II from hospital - stored data. Method 7623 patient specimens of 47 the Renmin Hospital of Wuhan University were collected. Indirect reference intervals 48 were established based on the hospital - stored data with four different methods, 49 including the Hoffmann method (HM), revised Hoffmann method (HMCDF), E - M 50 algorithm - based method (EMBCT), and a recent estimator (KOSMIC). According to 51 CLSI C28 - A3 guidelines, 369 healthy specimens were collected. The authors tested 52 the difference between reference intervals of gender - specific and age - specific 53 subgroups using Harris and Boyd's test. Finally, the averaging result of estimates was 54 calculated according to how likely each model was. Results The indirect reference 55 intervals of PIVKA - II based on LIS data were 0 to 35.30 mAU/mL (HM), 0 to 31.48 56 mAU/mL (HMCDF), 0 to 30.78 mAU/mL (EMBCT), 0 to 36.17 mAU/mL (KOSMIC) 57 and 0 to 31.48 mAU/mL (averaging) respectively, and the reference intervals based on 58 healthy group were 0 to 32 mAU/mL. Compared with HM, EMBCT and KOSMIC, 59 HMCDF and the averaging result was closer to those of the health group. Significant 60 difference was detected between gender - partitioned subgroups, and the reference 61 upper limit in the female group was smaller than the male group. Conclusions The 62 authors established the indirect reference intervals of PIVKA - II for the Wuhan 63 population, which could be used to the clinical reference intervals. The framework 64 proposed could help clinical laboratory set their reference intervals of test items.

Human abnormal prothrombinogen (PIVKA-II), a protein induced by vitamin K 81 deficiency or antagonist II, is also known as dextro-γ-carboxy-prothrombinogen. 82 Published research suggests that patients may have abnormal liver metabolism due to 83 vitamin K deficiency, allowing incomplete carboxylation of several glutamates near 84 the amino terminus of prothrombinogen to form abnormal prothrombin and lose the 85 related coagulation activity 4,5 . In HCC patients, the endoplasmic reticulum can not 86 carboxylate PIVKA-II to become active normal prothrombin, resulting in elevated 87 serum PIVKA-II levels. If PIVKA-II was found to be abnormally elevated in serum, it might provide a basis for detecting HCC to some extent. The literature suggests that 89 PIVKA-II is elevated in a certain percentage of HCC patients 6 and has high diagnostic 90 specificity, especially in AFP-negative hepatocellular carcinoma patients. It has been 91 suggested that PIVKA-II is generally more effective than AFP and has an excellent 92 clinical application in combination with AFP in tandem or a parallel inspection 7 .

93
Due to insufficient literature and research related to PIVKA-II in China, many clinical 94 laboratories generally follow two ideas for determining its reference interval. One is 95 to cite the literature or the manufacturers' reagent instructions. The second is to 96 transfer and verify its biological reference interval. Due to race, age, and gender 97 differences and the influence of testing methods, instruments, and reagents, the 98 resulting reference intervals obviously cannot provide accurate clinical diagnosis and 99 treatment guidance. They may even cause trouble to clinical work and patients.

100
The American Clinical and Laboratory Standardization Institute (CLSI) document 101 C28-A3 is currently widely recognized, which recommends that each clinical 102 laboratory establish reference intervals appropriate to its condition 8 . However, setting 103 reference intervals following the CLSI files takes a lot of time and money. Moreover, 104 it is even more difficult in clinical work when reference intervals are needed to 105 differentiate age and genders, especially newborns and the elderly. Besides, for some 106 particular analytes, such as cerebrospinal fluid (CSF), it becomes tough to establish 107 reference intervals for tests in these situations 9 .

108
With computer information technology development, hospitals have gradually 109 established and improved their laboratory information systems (LIS), which store many test data and contain a massive amount of value yet to be explored. Several      Quality Control) were tested with the samples every working day. The instrument's 158 detection range was 5~75000 mAU/ml, with intra-batch precision ≤ 3.5% and 159 inter-batch precision ≤ 4.5%.  The author partitioned the storage group data into gender-specific subgroups, and four 175 methods were used to establish the indirect reference intervals for the subgroups.

186
where was the observed value and ( ) was the linear regression function.

187
The authors also determined the best-fitting piece-wise regression by least-squares In Figure 1, the authors found that the storage group data had a skewed distribution, interval, which assumed that the storage group data was sampled from two 207 distributions, the healthy and diseased distribution. The basic idea of the EMBCT was 208 to separate these two distributions. The observed PIVK-II testing data can be 209 derived as: Inserting the MLEs of and into the above equation, we found that ̂ became:  According to the non-parametric method recommended in CLSI document C28-A3,

307
PIVKA-II's original data contained 7623 cases, and the valid data were 1152 cases after three screenings, with a ratio of 15.11%. Details could be found in Table 1. HMCDF was the most probably model. Table 2, Table 3 and Table 4 detail the indirect 321 reference intervals of PIVKA-II based on the storage group data. In Table 5 The authors drew the cumulative frequency distribution chart and curve of PIVKA-II 338 for the healthy group and gender-specific subgroups in Figure 1. PIVKA-II's 95% 339 right-sided reference interval was 0~32.00 mAU/mL, and the 90% PIVKA-II 340 reference upper limit confidence interval was 31.00~32.60 mAU/mL. Age-specific 341 reference intervals, gender-specific reference intervals, the median and IQR 342 (interquartile range) were calculated in Table 6. improved, and a relatively mature methodological system has been established 20-22 .

363
As early as 1963, some researchers established serum glucose reference intervals by 364 Hoffmann method 11 , which screened hospital mixed data and made an initial attempt 365 to establish reference intervals with hospital storage data. Then the value of hospital 366 storage data was gradually discovered.

367
This study used four methods to calculate PIVKA-II's indirect reference intervals 368 from LIS's stored data.

369
The first method was the Hoffmann method, which determined the linear portion with

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.